CN109917656A - Recirculated cooling water minimum differntial pressure energy-saving control system and method based on processing medium multi-temperature target - Google Patents
Recirculated cooling water minimum differntial pressure energy-saving control system and method based on processing medium multi-temperature target Download PDFInfo
- Publication number
- CN109917656A CN109917656A CN201910251616.9A CN201910251616A CN109917656A CN 109917656 A CN109917656 A CN 109917656A CN 201910251616 A CN201910251616 A CN 201910251616A CN 109917656 A CN109917656 A CN 109917656A
- Authority
- CN
- China
- Prior art keywords
- processing medium
- cooling
- medium temperature
- value
- temperature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000012545 processing Methods 0.000 title claims abstract description 262
- 239000000498 cooling water Substances 0.000 title claims abstract description 109
- 238000000034 method Methods 0.000 title claims description 60
- 238000001816 cooling Methods 0.000 claims abstract description 174
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 168
- 230000003134 recirculating effect Effects 0.000 claims abstract description 52
- 230000002349 favourable effect Effects 0.000 claims abstract description 45
- 238000012549 training Methods 0.000 claims description 54
- 230000008859 change Effects 0.000 claims description 41
- 239000011159 matrix material Substances 0.000 claims description 34
- 238000001514 detection method Methods 0.000 claims description 27
- 238000013528 artificial neural network Methods 0.000 claims description 25
- 230000037452 priming Effects 0.000 claims description 23
- 238000013135 deep learning Methods 0.000 claims description 18
- 238000005516 engineering process Methods 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 14
- 238000005070 sampling Methods 0.000 claims description 6
- 238000012216 screening Methods 0.000 claims description 6
- 210000005036 nerve Anatomy 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 5
- 238000004134 energy conservation Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims 1
- 230000008901 benefit Effects 0.000 abstract description 2
- 238000004519 manufacturing process Methods 0.000 description 18
- 230000006870 function Effects 0.000 description 11
- 238000009776 industrial production Methods 0.000 description 9
- 230000000694 effects Effects 0.000 description 8
- 210000002569 neuron Anatomy 0.000 description 7
- 239000000126 substance Substances 0.000 description 7
- 230000004913 activation Effects 0.000 description 6
- 238000013461 design Methods 0.000 description 6
- 238000005457 optimization Methods 0.000 description 6
- 241000196324 Embryophyta Species 0.000 description 5
- 230000001276 controlling effect Effects 0.000 description 5
- 230000015572 biosynthetic process Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000003786 synthesis reaction Methods 0.000 description 4
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000005265 energy consumption Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 239000008400 supply water Substances 0.000 description 3
- 239000002699 waste material Substances 0.000 description 3
- XSQUKJJJFZCRTK-UHFFFAOYSA-N Urea Chemical compound NC(N)=O XSQUKJJJFZCRTK-UHFFFAOYSA-N 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 239000004202 carbamide Substances 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012804 iterative process Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 1
- 240000002853 Nelumbo nucifera Species 0.000 description 1
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000005272 metallurgy Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005086 pumping Methods 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 238000005057 refrigeration Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 101150027113 sigM gene Proteins 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P80/00—Climate change mitigation technologies for sector-wide applications
- Y02P80/10—Efficient use of energy, e.g. using compressed air or pressurized fluid as energy carrier
Abstract
The invention discloses a kind of recirculated cooling water minimum differntial pressure energy-saving control systems based on processing medium multi-temperature target, including recirculating cooling water system and control energy conserving system；Control energy conserving system includes processing medium temperature least favorable point selector, processing medium temperature controller and gives back to differential water pressures inner ring PID controller；Processing medium temperature least favorable point selector is for selecting processing medium temperature least favorable point；Processing medium temperature controller gives back to differential water pressures setting value for obtaining cooling；Differential water pressures inner ring PID controller is given back to for controlling the aperture of all upper tower valves and the outlet stream amount of water supply pump assembly.The utility model has the advantages that the cooling supply amount of recirculated cooling water can be changed automatically according to industry spot processing medium cooling demand by so that recirculated cooling water is not required to manual adjustment, recirculating cooling water system energy utilization rate is improved, realizes Energy Saving Control.
Description
Technical field
The present invention relates to recirculated cooling water technical fields in industrial production, specifically a kind of how warm based on processing medium
Spend the recirculated cooling water minimum differntial pressure energy-saving control system and method for target.
Background technique
In the industrial processes such as chemical industry, electric power, metallurgy, system is as needed, often because of burning, chemical reaction
The equal a large amount of heat of generations causes equipment and system temperature to increase, influences production district processing medium quality, cause serious economy
Loss.Recirculating cooling water system is a kind of common temperature control engineering system in industrial site to processing medium, utilizes biography
The heat generated in production process is transmitted in natural environment by thermal medium, achievees the purpose that cooling, is widely used.Circulating cooling
Water system network topology is in large scale, and structure is complicated, and its internal component is various, and system design is to rely on experience mostly,
Needs are produced to meet, usual foundation peak load simultaneously gives certain surplus capacity and blindly improves water supply capacity and cooling capacity,
It is commonly present " low load with strong power " phenomenon between actual industrial production requirement, causes a large amount of cooling resource waste.
Traditional industrial circulating cooling water system uses more constant speed water pump parallel pumpings, when industrial production area is to cooling water
When demand changes, the outlet valve by adjusting each water pump controls cooling water flow.It but is anti-at actual industrial scene
Only opening big exit of pump valve causes pump motor electric current to increase and damage pump motor, or because only turning down valve with artificial experience
Reduce water flow and influence production district cooling effect, the outlet valve aperture of each water pump is often set as 50%, is entirely being transported
The row period causes great restriction loss, without fundamentally solving the problems, such as that cooling resource wastes.Water pump is as circulating cooling
The main energy consumption equipment of water system, electric quantity consumption is surprising, a recirculating cooling water system 1 year large-scale pump working power consumption
Up to ten million yuan, therefore pump performance parameters and operating condition will affect the economical and energy saving effect of whole system.It is supplying water in China
Water pump operation management level is relatively low in pipe network system, and in most cases the actual condition of water pump operation can deviate design
Operating condition, so that water pump efficiency substantially reduces.In document, " Zheng Lei, explains expensive, and Li Tongling water pump vane cuts new technology
2005,31 (9): practice and exploration [J] water supply and drainage propose the impeller outer diameter by cutting water pump to adjust water in 94-96. "
Flow, but this method performance difficulty, and flow adjustment range is limited, cannot follow the variation of flow demand and be adjusted on a large scale
Section.In document, " Fu Yongzheng, Cai Yaqiao closed-loop water system multiple ontology become the flow rate calculation and prediction [J] that number of units is adjusted
2007 (3): pump technology proposes that changing parallel water pump runs number of units, realizes that flow is adjusted, this method principle letter in 23-26. "
It is single, but degree of regulation is low, cannot flow is finely adjusted or be continuously adjusted.In document " Hickok H N.Adjustable
Speed---A Tool for Saving Energy Losses in PuM’ps,Fans,Blowers,and CoM’
pressors[J].IEEE Transactions on Industry Applications,1985,IA-21(1):124-
It is proposed in 136. " to change pump rotary speed regulating water flow using variable-frequency control technique, this method adjustable range is wider, precision compared with
Height does not increase pipe resistance additionally, and water pump can be made to keep efficient operation, can be fundamentally energy saving.Variable-frequency control technique and control
Technology processed combines, and applies to recirculated cooling water pump motor rotational speed regulation, constitutes recirculated cooling water pump variable frequency for water management system
System.
But the setting of industrial heat exchange device is highly and different to the distance of service water outlet, wants to water pump pressure of supply water
Ask also different, due to heat exchanger installation site be it is fixed, there are a fixed pressure in water supply network least
Benefit point requires highest to water pump pressure of supply water.To enable cooling water to reach each heat exchanger with certain pressure, by right
The pressure of water supply network pressure least favorable point carries out tracing control, and variable frequency adjustment pump rotary speed keeps the point pressure constant, maintains
Whole system hydraulic equilibrium, this is current most widely used recirculated cooling water pump variable frequency water supply control program.In the program
The outlet pressure of water pump is changed with water consumption, because of referred to herein as recirculated cooling water pump variable frequency variable-pressure water supply control system.
The transformation of recirculated cooling water pump variable frequency has very protection circulation-water pump electric machine and raising network load power factor
Big change, but water circulating pump energy saving in running rate is not still high.This is because in design cycle cooling water pump variable frequency variable-pressure water supply
Following two factor is not accounted for when system:
First, industrial production environment influences cooling water temperature, autumn and winter environment temperature is obviously than spring and summer season environment temperature
Spend low, night environment temperature and daytime environment temperature also have the biggish temperature difference.Industrial production environment is to recirculating cooling water system
Heat exchange influences specific manifestation are as follows: and industrial production environment temperature is lower, and the cooling capacity of cooling tower in circulating cooling water system is stronger,
The feed temperature to cool down by cooling tower is lower, and the cooling water of identical water flow is brighter to the cooling effect of same heat exchange load
It is aobvious, that is, it is directed to identical heat exchange demand, the circulating cooling water flow needed is smaller.
Second, industrial heat exchange load variations influence cooling requirement amount, the industrial heat exchange load in dull season is produced obviously than production
Busy season is small, and after industrial production quotient receives rush order increasing output, heat exchange load is obviously bigger than daily production.Industrial heat exchange is negative
Lotus, which changes the heat exchange on recirculating cooling water system, influences specific manifestation are as follows: wants to ensure that industrial production meets normal production always
It asks, industrial heat exchange load is bigger, and the amount of cooling water needed is bigger, for bigger with mutually synthermal cooling water inflow demand.
Summary of the invention
In view of the above-mentioned problems, the present invention provides a kind of recirculated cooling water minimum pressures based on processing medium multi-temperature target
Poor energy-saving control system and method.Change the cooling supply of recirculated cooling water automatically according to industry spot processing medium cooling demand
Amount, realize Energy Saving Control.In order to achieve the above objectives, the specific technical solution that the present invention uses is as follows:
A kind of recirculated cooling water minimum differntial pressure energy-saving control system based on processing medium multi-temperature target, including circulation are cold
But water system and control energy conserving system；The recirculating cooling water system includes the N number of cooling tower being arranged along circulation waterway, water suction
Pond, water supply pump assembly, outlet pipe group, a heat exchanger of M ' and return pipe group；Cooling bay is provided in cooling tower；The water outlet
Pipe group includes N root priming reservoir water inlet pipe, L root priming reservoir outlet pipe, water main, M ' root subfeeder；The return pipe group packet
Include M ' root return branch, return main, tower return pipe on N root；Any cooling tower is corresponding to intake through a priming reservoir
Pipe is connect with the priming reservoir, and the priming reservoir is connect through priming reservoir outlet pipe described in L root with the water main side by side, described
Water main is corresponded to a heat exchanger of M ' through M ' root subfeeder and is supplied water；It has been arranged in parallel on the priming reservoir outlet pipe
Feed pump；Any heat exchanger correspondence is connect through a return branch with the return main, and the return main is through arranged side by side
N root on tower return pipe connect one to one with N number of cooling tower；Processing medium temperature biography is provided in a heat exchanger of M '
Sensor, the processing medium temperature sensor is for obtaining various processing medium real time temperature detected values；
Cooling feed temperature sensor and cooling feed pressure sensor, the cooling are provided on the water main
Feed temperature sensor for obtaining cooling feed temperature detected value, the cooling feed pressure sensor for obtain it is cooling to
Water pressure detected value；Cooling backwater temperature sensor and cooling backwater pressure sensor, institute are provided on the return main
Cooling backwater temperature sensor is stated for obtaining cooling backwater temperature detection value, the cooling backwater pressure sensor is for obtaining
Cooling backwater pressure detection value；A upper tower valve is respectively arranged on the upper tower return pipe described in N root；
Its key technology is: including processing medium temperature least favorable point selector, technique in the control energy conserving system
Medium temperature controller and give back to differential water pressures inner ring PID controller；The processing medium temperature least favorable point selector is according to institute
The collected processing medium temperature deviation value sequence of recirculating cooling water system is stated, processing medium temperature least favorable point is obtained；It is described
Processing medium temperature controller according to the processing medium temperature least favorable point, and then in conjunction with field operational data obtain it is cooling to
Return water pressure differential resetting value；
The cooling feed pressure detected value and cooling backwater pressure detection value that the recirculating cooling water system detects are made
Cooling is obtained after difference gives back to differential water pressures detected value；The cooling give back to differential water pressures detected value and it is cooling give back to differential water pressures setting value make it is poor
The cooling obtained afterwards gives back to differential water pressures deviation；It is described to give back to differential water pressures inner ring PID controller to give back to differential water pressures inclined according to cooling
Difference adjusts the apertures of all upper tower valves, to change the outlet stream amount of water supply pump assembly.
By above-mentioned design, it on the basis of recirculating cooling water system, is controlled in conjunction with pump variable frequency variable-pressure water supply, establishes base
In the recirculated cooling water minimum differntial pressure energy-saving control system of processing medium multi-temperature target.By the online temperature detection value of processing medium
It is known as processing medium temperature deviation with the difference of its desired temperature, each processing medium temperature deviation is minimum in entire heat exchanger group
Point is known as processing medium temperature least favorable point.This recirculating cooling water system is according to the temperature deviation of processing medium temperature least favorable point
And deviation variation rate, in conjunction with cooling feed temperature real-time detection value, control all processing medium temperature of industry spot are maintained
In normal safe and energy-efficient production temperature range, the demand of cooling water flow is determined in real time, and tower valve in adjusting changes water pump
Cooling water outlet amount.Finally, on the basis of the core requirement that satisfaction controls production technology medium temperature, fundamentally into one
Step realizes pump motor energy conservation in recirculating cooling water system.
Further technical solution are as follows: the processing medium temperature deviation value sequence includes a processing medium temperature deviation of M '
Value, any processing medium temperature deviation value are equal to the corresponding processing medium real time temperature detected value and the corresponding technique
The difference of medium temperature setting value；The processing medium temperature deviation change rate sequence be include a processing medium temperature deviation of M '
Change rate, wherein any processing medium temperature deviation change rate is the corresponding processing medium two neighboring detection period
The ratio of temperature change value and a upper detection period.
Using the above scheme, in conjunction with recirculating cooling water system, processing medium temperature acquisition is realized, in conjunction with detected value and setting
Value, obtains the difference in all processing mediums, the heat exchanger and processing medium for finding place closest to temperature given threshold, comes
Upper tower valve control is carried out to whole system, guarantees that each processing medium is maintained within the scope of safety and energy efficiency temperature.
A kind of control method of the recirculated cooling water minimum differntial pressure energy-saving control system based on processing medium multi-temperature target,
It is characterized in that specific steps are as follows:
S1: the setting sampling period carries out field operational data acquisition to the recirculating cooling water system；
The recirculating cooling water system passes through in a heat exchanger of processing medium temperature sensor collection in worksite M '
A processing medium real time temperature detected value of M '；The recirculating cooling water system passes through the cooling feed temperature sensor scene
Acquire cooling feed temperature detected value；The recirculating cooling water system is cold by the cooling feed pressure sensor collection in worksite
But feed pressure detected value；The recirculating cooling water system passes through the cooling backwater temperature sensor collection in worksite cooling backwater
Temperature detection value；The recirculating cooling water system is examined by the cooling backwater pressure sensor collection in worksite cooling backwater pressure
Measured value；
S2: according to a processing medium real time temperature detected value of the M ' of collection in worksite and corresponding a processing medium temperature of M '
Setting value obtains a processing medium temperature deviation value of M ' and corresponding a processing medium temperature deviation change rate of M '；M ' is a described
Processing medium temperature deviation value constitutes the processing medium temperature deviation value sequence；A processing medium temperature deviation rate of change value of M '
Constitute the processing medium temperature deviation change rate sequence；
S3: the temperature in the processing medium temperature least favorable point selector selection processing medium temperature deviation value sequence is inclined
Poor minimum value obtains the corresponding processing medium temperature of processing medium temperature least favorable point as processing medium temperature least favorable point
Spend deviation variation rate and corresponding heat exchanger；
S4: processing medium temperature controller goes to obtain corresponding heat exchanger according to the processing medium temperature least favorable point
The field operational data of corresponding heat exchanger is inputted and is corresponded to by processing medium temperature deviation value, processing medium temperature deviation change rate
Processing medium temperature prediction model obtains cooling and gives back to differential water pressures setting value and processing medium temperature prediction value；
S5: the cooling feed pressure detected value and cooling backwater pressure detection value are obtained into cooling as difference and give back to differential water pressures
Detected value；The cooling gives back to differential water pressures setting value and the cooling gives back to after differential water pressures detected value makees difference and obtains cooling to return water
Differential pressure tolerance value, which, which gives back to, gives back to the upper tower valve of differential water pressures inner ring PID controller adjusting described in differential water pressures deviation feeding
Aperture, to change the outlet stream amount with water supply pump assembly.
The heat exchanger water temperature control closed loop of above-mentioned method, production district supervises the cooling situation of each production district to be cooled in real time
It surveys, and the aperture for adjusting tower valve keeps required cooling water inflow in each production district heat exchanger to be cooled precisely controlled, is inciting somebody to action
Cooling resource waste is reduced while the control of processing medium temperature is in range of keeping the safety in production, and is also cooled down while realizing energy conservation cold
But effect is more preferable.When upper tower valve opening, which reaches maximum, to control effectively to processing medium temperature, it can take and add
The measure of big cooling tower refrigeration effect.
Further, processing medium temperature deviation is found out using processing medium temperature least favorable point selector in step S3
Specific steps of the minimum value as processing medium temperature least favorable point in value sequence are as follows:
S31: initialization amounts to a processing medium of M ' if a processing medium temperature deviation value of M ' forms a difference group
Temperature deviation value, enables Wk=M '；K=1
S32: W is enabledk+1=Wk+X, makes Wk+1It can be divided exactly by M ', X is the wireless big vacancy of difference of filling；And X is equal to 0
~M ' -1；
S33: W is calculatedk+2=Wk+1/M’
S34: from Wk+2In group, using comparison method is intersected, found out most from a processing medium temperature deviation value of each group of M '
Small value, obtains Wk+2A processing medium temperature deviation value；
S35: judge Wk+2Whether 1 is equal to；If so, least using the processing medium temperature deviation value as processing medium temperature
Sharp point；Otherwise, k=k+2 is enabled；Return step S32.
It is additionally arranged least favorable point selector, finds the processing medium temperature deviation minimum value of big quantity, to determine that technique is situated between
Matter temperature least favorable point focuses on the real-time and accuracy of the selection of processing medium least favorable point.
Further, the foundation of a processing medium temperature prediction model deep learning neural network of M ' in step S4 walks
Suddenly are as follows:
S411: being numbered a heat exchanger of M ', and obtains recirculating cooling water system and run generation within X sampling period
Historical data, and the historical data that will acquire, after being screened according to screening conditions, as processing medium temperature prediction in heat exchanger
The training data of model；
S412: determining characteristic variable from historical data, and the processing medium temperature in the historical data of heat exchanger I is inclined
Difference, processing medium temperature deviation change rate, cooling feed temperature detected value, the cooling differential water pressures detected value that gives back to are as input number
According to, and normalizing data set is obtained after carrying out data normalized, and the normalizing data set is divided into training sample set and survey
Try sample set；
S413: based on autocoder is stacked, successively greedy unsupervised pre-training is carried out to training sample set, obtains depth
The input layer of learning neural network initialization and weight matrix W, input layer and the hidden layer threshold value matrix B of hidden layer；
S414: small parameter perturbations are carried out: the weight matrix of input layer and hidden layer to the initialization of deep learning neural network
W, input layer and hidden layer threshold value matrix B are finely adjusted, and until the number of iterations reaches the number of iterations maximum value, are based on
Stack the initial process medium temperature prediction model of autocoder；
S415: the initial process medium temperature prediction model that step S414 is obtained is commented using test sample data set
Estimate, obtains based on the processing medium temperature prediction model for stacking autocoder.
Based on the processing medium temperature prediction model for stacking autocoder, compared with traditional shallow-layer neural network, heap
Stack autocoder efficiently solves a series of problems caused by traditional neural network stochastic parameter initializes, and can effectively dig
Each data implication relation is dug, industrial site processing medium temperature prediction precise control is greatly improved；Compared to biography
Industrial site processing medium temperature prediction control precision can be improved in system storehouse autocoder method.By this method application
In industrial site recirculating cooling water system Energy Saving Control, it can quick and precisely provide cooling and give back to differential water pressures setting value, have
Processing medium temperature changing trend in each heat exchanger is grasped in real time conducive to industrial site administrative staff, it is most important that is reduced
Recirculating cooling water system energy consumption.
Further, the characteristic variable includes processing medium temperature deviation value, the technique in any one heat exchanger
Cooling feed temperature inspection in medium temperature deviation variation rate, processing medium real time temperature detected value and recirculating cooling water system
Measured value, cooling give back to differential water pressures detected value.
Further, the screening conditions are that processing medium temperature is in going through in safe and energy-efficient temperature value section
History data, described safe and energy-efficient temperature value section is in processing medium temperature threshold section.
Difference between safe and energy-efficient temperature value section and processing medium temperature threshold section, according to technical staff's
Historical experience technology is set after being concluded.
Further, in step S413, when carrying out successively greedy unsupervised pre-training to training sample set, by training sample
This collection is divided into P group small lot training sample, is successively trained, and uses Dropout technology, randomly selects partial nerve member temporarily
It stops work and makees, successively iteration, successively training is the weight matrix W of the input layer of deep learning neural network initialization and hidden layer, defeated
Enter layer and hidden layer threshold value matrix B.
In step S413, stacking autocoder SAE (stacked autoencoder) is a kind of typical deep learning
Neural network, basic Component units are autocoder (antoencoder, AE), and network structure is by encoder and decoding
Device composition: being mapped as the feature vector in hidden layer for input vector by encoder, then passes through decoder for proper phasor
It is reconstructed into original input vector.
As a given input sample set X={ xi| 1≤i≤N }, wherein N is sample total number, xiFor in sample set
I-th of training sample, dimension n.If H={ hi| 1≤i≤N } it is hidden layer feature vector set, hiFor i-th of sample pair
The feature vector answered, dimension M ', the then encoding relation of X and H are as follows:
H=sf(WX+B)
In formula: W is the weight matrix of input layer and hidden layer；B is input layer and hidden layer threshold value matrix；sfFor encoder
Neuron activation functions, generally use Sigmoid function, with good feature identification degree:
sf(z)=1/ (1+exp (- z))
In formula: z is input vector.
Decoder is the inverse operation of encoder, using the feature vector of hidden layer as input vector, if
For output vector set,For the corresponding output vector of i-th of sample, dimension n, then the expression formula of decoder are as follows:
In formula: W ' is the weight matrix of hidden layer and output layer；B ' is the threshold matrix of hidden layer and output layer；Sg is solution
The neuron activation functions of code device.Autocoder is reached by minimizing the reconstructed error between output vector and input vector
Formula to the purpose of feature extraction, reconstructed error is as follows:It is constantly adjusted using gradient descent algorithm
Whole network weight and threshold value reduce reconstructed error, and formula is as follows:
In formula: l is learning rate；It indicatesLocal derviation is asked to weight W；It indicatesTo threshold
Value B seeks local derviation.
SAE is a kind of deep learning neural network model being made of AE stack stacking, and the output of lower layer AE will be used as upper
The input of layer AE.Gradually being abstracted for feature is realized by the stacking of AE, ultimately forms more compact, useful feature.WnIt is
The weight matrix of n-1 layers of hidden layer and n-th layer hidden layer, BnFor the threshold matrix of (n-1)th layer of hidden layer and n-th layer hidden layer.
Training process is divided into greedy successively unsupervised pre-training and has supervision two steps of fine tuning.Greedy successively unsupervised pre-training passes through
Successively training obtains the initialization weight and threshold value of network, and the input of bottom AE is initial data, hidden layer output data conduct
The input data of upper layer AE.
After the completion of being layered pre-training, hidden layer is stacked, input data is expressed as with output data relationship:
In formula: f is activation primitive, xiFor the input variable of i-th of training sample, W, B are respectively that layer-by-layer pre-training obtains
Netinit weight and threshold value,For the predicted value of i-th of sample.The error for constructing actual value and predicted value loses letter
Number, formula are as follows:
In formula: N is sample total number, yiFor the actual value of i-th of sample.By top-down backpropagation to entire
Network weight and threshold value are finely adjusted, and reduce the error of predicted value and actual value.
SAE can effectively extract the high-order feature of data by successively greedy unsupervised pre-training, preferably approach complexity
Function, and parameter optimization space is reduced, network parameter can be quickly obtained, the further feature learning ability of neural network is improved.
Compared to shallow-layer machine learning algorithm, deep layer network is when handling high dimensional data, because its complicated network structure is more
Add and be easy to produce overfitting problem, thus the generalization ability of limited model.
Dropout is a kind of anti-over-fitting technology of mainstream, basic thought are as follows: one is randomly choosed in model training
Partial node does not work, these nodes will save the weight of last iteration, and output is set to 0.These nodes selected exist
The weight retained can restore during next iteration again before randomly chooses part of nodes again and repeats this process.Network knot
All certain variation will occur for structure in each iterative process, using Dropout technology, randomly select partial nerve member temporarily not
Work, reduces the collective effect between specific node, dependence of the network output to specific node state is alleviated, to prevent
Over-fitting.
Further, it when assessing initial process medium temperature prediction model, is surveyed using test sample data set
The improvement storehouse autocoder of examination after training is compiled using mean percent ratio error (MAPE) as improvement storehouse is measured automatically
The standard of code device assessment performance, expression formula are as follows:
In formula: yi、Respectively i-th of sample loops cooling gives back to the actual value of differential water pressures and assesses to obtain by SAE
Predicted value.In step S414, when small parameter perturbations, using top-down small lot RM ' SProp optimization method to deep learning
The weight matrix W threshold matrix B of neural network initialization is finely adjusted.
The specific steps of which are as follows: global learning rate l, rate of decay ρ is arranged, cumulative variations r1=0, r2=0 are initialized.
The small lot data set comprising a sample of M ' is chosen from training set, according to error loss function, calculates gradient:
Accumulative squared gradient is calculated, as shown in formula (9):
In formula: ⊙ is by element product symbol.
Weight and threshold parameter are updated respectively:
When the number of iterations reaches requirement, stop operation, otherwise returns to step 2 and continue to execute calculating.
Further describe, given back to described in step S5 differential water pressures inner ring PID controller to the aperture of the upper tower valve and
The specific control content of the outlet stream amount of the water supply pump assembly are as follows:
If the cooling differential water pressures deviation value P that gives back to, greater than 0, the differential water pressures inner ring PID controller that gives back to issues upper tower valve tune
Big control signal increases valve opening for controlling upper tower valve；If the cooling differential water pressures deviation value P that gives back to is equal to 0, described to give return water
Pressure difference inner ring PID controller keeps former state of a control；
If cooling give back to differential water pressures deviation value P less than 0, the differential water pressures inner ring PID controller that gives back to issues upper tower valve tune
Small control signal reduces valve opening for controlling upper tower valve；By above-mentioned design, it realizes and gives back to differential water pressures deviation in conjunction with cooling
Value, is adjusted the aperture of upper tower valve, to allow all processing medium temperature to guarantee in threshold value.And guarantee the feature of environmental protection
Energy.
Beneficial effects of the present invention: so that recirculated cooling water is not required to manual adjustment can cool down according to industry spot processing medium
Demand changes the cooling supply amount of recirculated cooling water automatically, improves recirculating cooling water system energy utilization rate, realizes Energy Saving Control.
Detailed description of the invention
Fig. 1 is chemical plant A recirculating cooling water system artwork；
Fig. 2 is chemical plant A feed pump unit frequency-conversion variable-pressure water supply control block diagram；
Fig. 3 is water synthesis zone heat exchanger group schematic diagram in the A recirculating cooling water system of chemical plant；
Fig. 4 is that processing medium temperature least favorable point finds flow chart；
Fig. 5 is the recirculated cooling water minimum differntial pressure energy-saving control system control block diagram based on processing medium multi-temperature target；
Fig. 6 is the recirculated cooling water minimum differntial pressure Energy Saving Control flow chart based on processing medium multi-temperature target；
Fig. 7 is processing medium multi-temperature target set point switching many reference amounts predictive control algorithm flow chart；
Fig. 8 is autocoder structure chart；
Fig. 9 is to stack self-encoding encoder framework；
Figure 10 is Dropout self-encoding encoder framework.
Specific embodiment
Specific embodiment and working principle of the present invention will be described in further detail with reference to the accompanying drawing.
A kind of recirculated cooling water minimum differntial pressure energy-saving control system based on processing medium multi-temperature target, including circulation are cold
But water system and control energy conserving system.
In the present embodiment, by taking the A of chemical plant as an example, recirculated cooling water Energy Saving Control is carried out.It will be seen from figure 1 that chemical industry
Factory A recirculating cooling water system artwork, in conjunction with Fig. 1 as can be seen that the recirculating cooling water system includes setting along circulation waterway
N number of cooling tower, priming reservoir, water supply pump assembly, outlet pipe group, a heat exchanger of M ' and the return pipe group set；It is set in cooling tower
It is equipped with cooling bay；In the present embodiment, M ' is positive integer, and there are three heat exchanger groups: synthesis zone heat exchanger group, urea area heat exchanger
Group, power-section heat exchanger group.
The outlet pipe group includes N root priming reservoir water inlet pipe, L root priming reservoir outlet pipe, water main, M ' root water supply branch
Pipe；In the present embodiment, heat exchanger quantity M '=50.Wherein, 50 heat exchangers are divided into three heat exchanger groups, respectively synthesis zone
Heat exchanger group, urea go heat exchanger group, power to go heat exchanger group.Three heat exchanger groups amount to 50 heat exchangers.In the present embodiment
In, the return pipe group includes M ' root return branch, return main, tower return pipe on N root；
In the present embodiment, any cooling tower is corresponding connects through a priming reservoir water inlet pipe and the priming reservoir
It connects, the priming reservoir, is connect side by side through priming reservoir outlet pipe described in L root with the water main, the water main is through M ' root
Subfeeder is corresponded to a heat exchanger of M ' and is supplied water；Feed pump has been arranged in parallel on the priming reservoir outlet pipe；
In the present embodiment, any heat exchanger correspondence is connect through a return branch with the return main, described
Return main connects one to one through tower return pipe on N root arranged side by side with N number of cooling tower；
In the present embodiment, it is provided with processing medium temperature sensor in a heat exchanger of M ', the processing medium temperature
Degree sensor is for obtaining various processing medium real time temperature detected values；
In the present embodiment, it is provided with cooling feed temperature sensor on the water main and cooling feed pressure passes
Sensor, the cooling feed temperature sensor is for obtaining cooling feed temperature detected value, the cooling feed pressure sensor
For obtaining cooling feed pressure detected value；
In the present embodiment, it is provided with cooling backwater temperature sensor on the return main and cooling backwater pressure passes
Sensor, the cooling backwater temperature sensor is for obtaining cooling backwater temperature detection value, the cooling backwater pressure sensor
For obtaining cooling backwater pressure detection value；
In the present embodiment, a upper tower valve is respectively arranged on the upper tower return pipe described in N root；
In the present embodiment, can be seen that in conjunction with Fig. 2, Fig. 3 and Fig. 5 in the control energy conserving system includes processing medium
Temperature least favorable point selector, processing medium temperature controller and give back to differential water pressures inner ring PID controller；The processing medium temperature
Least favorable point selector is spent according to the collected processing medium temperature deviation value sequence of the recirculating cooling water system, obtains technique
Medium temperature least favorable point；
The processing medium temperature controller combines scene operation number according to the processing medium temperature least favorable point
Differential water pressures setting value is given back to according to cooling is obtained；
The cooling feed pressure detected value and cooling backwater pressure detection value that the recirculating cooling water system detects are made
Cooling is obtained after difference gives back to differential water pressures detected value；The cooling give back to differential water pressures detected value and it is cooling give back to differential water pressures setting value make it is poor
The cooling obtained afterwards gives back to differential water pressures deviation；It is described give back to differential water pressures inner ring PID controller cooling give back to differential water pressures deviation
The aperture for controlling all upper tower valves is adjusted, to change the outlet stream amount of water supply pump assembly.
In conjunction with Fig. 3, synthesis zone heat exchanger group schematic diagram, cooling water supply enters in industrial production area by outlet pipe group is respectively changed
Hot device, there are temperature difference between the processing medium and cooling water supply in heat exchanger, processing medium constantly transfers heat to cooling
Itself temperature drop is realized in water supply, and cooling water supply, which carries heat exchange amount outflow heat exchanger group and send to return pipe group, becomes cooling backwater.Entirely
Heat transfer process must assure that different process medium temperature meets respective production requirement always in each heat exchanger, it is therefore desirable to industry
All processing mediums in production scene carry out real time temperature Detection & Controling.
As preferential, the processing medium temperature deviation value sequence includes a processing medium temperature deviation value of M ', according to changing
Hot device number successively calculates processing medium temperature deviation value according to number sorting and obtains processing medium temperature deviation value sequence.
Any processing medium temperature deviation value is equal to the corresponding processing medium real time temperature detected value and is situated between with the corresponding technique
The difference of matter desired temperature；
The processing medium temperature deviation change rate sequence be include a processing medium temperature deviation change rate of M ',
In, any processing medium temperature deviation change rate is the corresponding two neighboring detection period temperature change value of processing medium
With the ratio of a upper detection period.
In the present embodiment, processing medium temperature deviation value is corresponding with processing medium temperature deviation change rate.
In conjunction with Fig. 6 as can be seen that a kind of recirculated cooling water minimum differntial pressure energy conservation control based on processing medium multi-temperature target
The control method of system processed, it is characterised in that specific steps are as follows:
S1: the setting sampling period carries out field operational data acquisition to the recirculating cooling water system；
The recirculating cooling water system passes through in a heat exchanger of processing medium temperature sensor collection in worksite M '
A processing medium real time temperature detected value of M '；
Recirculating cooling water system passes through the cooling feed temperature detected value of cooling feed temperature sensor collection in worksite；
Recirculating cooling water system passes through the cooling feed pressure detected value of cooling feed pressure sensor collection in worksite；
Recirculating cooling water system passes through cooling backwater temperature sensor collection in worksite cooling backwater temperature detection value；
Recirculating cooling water system passes through cooling backwater pressure sensor collection in worksite cooling backwater pressure detection value；
S2: it is situated between according to a processing medium real time temperature detected value of the M ' of step S1 collection in worksite and a technique of corresponding M '
Matter desired temperature obtains a processing medium temperature deviation value of M ' and corresponding a processing medium temperature deviation change rate of M '；M'
A processing medium temperature deviation value constitutes processing medium temperature deviation value sequence；A processing medium temperature deviation of M ' becomes
Rate value constitutes processing medium temperature deviation change rate sequence；
S3: the temperature in the processing medium temperature least favorable point selector selection processing medium temperature deviation value sequence is inclined
Poor minimum value obtains the corresponding processing medium temperature of processing medium temperature least favorable point as processing medium temperature least favorable point
Spend deviation variation rate and corresponding heat exchanger；
S4: processing medium temperature controller goes the technique for obtaining corresponding heat exchanger according to processing medium temperature least favorable point
The field operational data of corresponding heat exchanger is inputted corresponding work by medium temperature deviation, processing medium temperature deviation change rate
Skill medium temperature prediction model obtains cooling and gives back to differential water pressures setting value and processing medium temperature prediction value；
S5: the cooling feed pressure detected value and cooling backwater pressure detection value are obtained into cooling as difference and give back to differential water pressures
Detected value；The cooling gives back to differential water pressures setting value and the cooling differential water pressures detected value that gives back to is made to obtain cooling after difference to give back to differential water pressures
Deviation, the cooling give back to and give back to differential water pressures inner ring PID controller described in differential water pressures deviation feeding, and tower valve opens in adjusting
Degree, to change the outlet stream amount of water supply pump assembly.In in this implementation, feed pump is centrifugal pump, in centrifugal pump in frequency conversion
Device connection.Differential water pressures inner ring PID controller is given back to described in step S5 and adjusts the aperture of upper tower valve, to change and water supply pump machine
The specific control content of the outlet stream amount of group are as follows:
If the cooling differential water pressures deviation value P that gives back to, greater than 0, the differential water pressures inner ring PID controller that gives back to issues upper tower valve tune
Big control signal increases valve opening for controlling upper tower valve；The differential water pressures inner ring PID controller that gives back to also issues feed pump
Machine set outlet water flow reduces control signal, for reducing water pump assembly outlet stream amount；If cooling gives back to differential water pressures deviation value P
Equal to 0, the differential water pressures inner ring PID controller that gives back to keeps former state of a control；If differential water pressures deviation value P is given back to less than 0 cooling,
The upper tower valve of differential water pressures inner ring PID controller sending that gives back to turns control signal down, reduces valve opening for controlling upper tower valve；
The differential water pressures inner ring PID controller that gives back to also is emitted to water pump assembly outlet stream amount increase control signal, for increasing water
Pump assembly outlet stream amount.As can be seen from Table I, chemical plant A recirculating cooling water system heat exchanger list, wherein work in table one
There is DSC computer screen temperature display point after skill media for heat exchange.
By each heat exchanger requirement listed in table, it is known that processing medium design objective and control range in each heat exchanger
It is different, therefore to ensure that all processing mediums maintain in respectively normal production temperature range, each technique should be detected one by one
Medium outlet temperature realizes that processing medium temperature integrally controls.Industrial site heat exchanger substantial amounts, processing medium type
It is various, in the corresponding desired temperature difference of each processing medium temperature detection value of line computation, the technique for constituting real-time update
Medium temperature biased sequence.If using upper tower valve regulation amount is determined again after calculating each processing medium cooling requirement amount
Method, control process is complicated and has larger redundancy, and will cause causes cooling water flow to adjust delay because computationally intensive, influences to control
Energy-saving effect processed.
Therefore the present invention proposes real-time selection and controls processing medium temperature least favorable point, processing medium temperature least favorable point
Refer to each processing medium detection temperature processing medium corresponding with the minimum value in the sequence of differences of set temperature, i.e. real-time control
Exceed the maximum processing medium of temperature setting range possibility in each heat exchanger, also with regard to the entire processing medium temperature of equivalent control
Degree.With industrial progress, processing medium temperature is constantly changing, and processing medium temperature least favorable point is real-time update
, therefore processing medium least favorable point temperature deviation and deviation variation rate are real-time as the input of processing medium temperature controller
It updates.
In conjunction with Fig. 4 as can be seen that finding out processing medium temperature using processing medium temperature least favorable point selector in step S3
Spend specific steps of the minimum value in deviation value sequence as processing medium temperature least favorable point are as follows:
S31: initialization amounts to a processing medium of M ' if a processing medium temperature deviation value of M ' forms a difference group
Temperature deviation value, enables Wk=M '；K=1；
S32: W is enabledk+1=Wk+X, makes Wk+1It can be divided exactly by M ', X is the wireless big vacancy of difference of filling；And X is equal to 0
~M ' -1；S33: W is calculatedk+2=Wk+1/M '
S34: from Wk+2In group, using comparison method is intersected, found out most from a processing medium temperature deviation value of each group of M '
Small value, obtains Wk+2A processing medium temperature deviation value；
S35: judge Wk+2Whether 1 is equal to；If so, least using the processing medium temperature deviation value as processing medium temperature
Sharp point；Otherwise, k=k+2 is enabled；Return step S32.
Further, in conjunction with Fig. 7 as can be seen that in step S4, the deep learning neural network of any heat exchanger
Establishment step are as follows: S411: recirculating cooling water system runs the historical data of generation within X sampling period, and what be will acquire goes through
History data, the training data after being screened according to screening conditions, as processing medium temperature prediction model in heat exchanger；Wherein, X is
Positive integer, in the present embodiment, X are equal to 1000.
S412: according to characteristic variable is determined from historical data, by the processing medium temperature in the historical data of heat exchanger I
Deviation, processing medium temperature deviation change rate, cooling feed temperature detected value, the cooling differential water pressures detected value that gives back to are as input number
According to, and normalizing data set is obtained after carrying out data normalized, and the normalizing data set is divided into training sample set and survey
Try sample set；
S413: based on autocoder is stacked, successively greedy unsupervised pre-training is carried out to training sample set, obtains depth
The input layer of learning neural network initialization and weight matrix W, input layer and the hidden layer threshold value matrix B of hidden layer；
S414: small parameter perturbations are carried out: the weight matrix of input layer and hidden layer to the initialization of deep learning neural network
W, input layer and hidden layer threshold value matrix B are finely adjusted, and until the number of iterations reaches the number of iterations maximum value, are based on
Stack the initial process medium temperature prediction model of autocoder；
S415: the initial process medium temperature prediction model that step S414 is obtained is commented using test sample data set
Estimate, obtains based on the processing medium temperature prediction model for stacking autocoder.
In conjunction with Fig. 7 as can be seen that the characteristic variable includes the processing medium temperature deviation value in any one heat exchanger
Ei, processing medium temperature deviation change rate Δ ei, in processing medium real time temperature detected value Tci and recirculating cooling water system
Cooling feed temperature detected value Tgs, cooling give back to differential water pressures detected value PΔj。
In the present embodiment, the screening conditions are that processing medium temperature is in safe and energy-efficient temperature value section
Historical data, described safe and energy-efficient temperature value section is in processing medium temperature threshold section.
In the present embodiment, the difference between described safe and energy-efficient temperature value section and processing medium temperature threshold section
Value is 4 DEG C.
In step S413, when carrying out successively greedy unsupervised pre-training to training sample set, training sample set is divided into P group
Small lot training sample, is successively trained, and uses Dropout technology, randomly selects partial nerve member break-off, successively
Iteration, successively training, the input layer and the weight matrix W of hidden layer of the initialization of deep learning neural network, input layer and implicit
Layer threshold matrix B.In step S414, when small parameter perturbations, using top-down small lot RMSProp optimization method to depth
Weight matrix W, input layer and the hidden layer threshold value matrix B of the input layer and hidden layer of practising neural network initialization are finely adjusted.
The pre-control algorithm is directly facing industrial circulating cooling water system field measurement data, is built respectively by deep layer framework
Founding each heat exchanger metric data includes processing medium temperature deviation and temperature deviation change rate, cooling feed temperature in each heat exchanger
Detected value and cooling give back to differential water pressures setting value, and with control amount premeasuring, i.e. cooling gives back to non-between differential water pressures setting value
Linear mapping relation.Using the two stages off-line training learning method of a kind of " pre-training-small parameter perturbations ", introduce simultaneously
Dropout technology and RMSPROP technology optimize processing medium temperature model parameters in each heat exchanger.Model after training
The stealth mode that deep structure mining data can be relied on, extracts the high-order for being conducive to processing medium temperature prediction control effect
Characteristic.In addition, this method can improve model generalization ability by the unsupervised training largely without mark sample.
In constructed SAE heat exchanger the number of plies of the hidden layer of the trained deep learning neural network of processing medium and
The neuron number of every layer of hidden layer has certain influence to Evaluation accuracy and off-line training time.With recirculating cooling water system scene
Detection data is successively configured hidden layer neuron number as sample input data: determining the 1st layer of hidden layer mind first
Optimal units and fixation through member, then increase the Optimal units of the 2nd layer of hidden layer neuron of one layer of determination, and so on, directly
Until mean percent ratio error (MAPE) no longer improves.
In order to keep technical solution of the present invention clearer, to the principle of storehouse autocoder used in the present invention into
Row is explained.SAE is a kind of typical deep learning neural network, and basic Component units are autocoders
(antoencoder, AE), network structure is as shown in Figure 8.Autocoder (autoencoder, AE) network structure such as Fig. 8
Shown, be made of encoder and decoder: then the feature vector being mapped as input vector by encoder in hidden layer is led to
It crosses decoder and proper phasor is reconstructed into original input vector.
As a given input sample set X={ xi| 1≤i≤N }, wherein N is sample total number, xiFor in sample set
I-th of training sample, dimension n.If H={ hi| 1≤i≤N } it is hidden layer feature vector set, hiFor i-th of sample pair
The feature vector answered, dimension M ', the then encoding relation of X and H are as follows:
H=sf(WX+B) in formula: W is the weight matrix of input layer and hidden layer；B is input layer and hidden layer threshold value square
Battle array；Sf is the neuron activation functions of encoder, generallys use sigM ' oid function, with good feature identification degree:
sf(z)=1/ (1+exp (- z))
In formula: z is input vector.
Decoder is the inverse operation of encoder, using the feature vector of hidden layer as input vector, if
For output vector set,For the corresponding output vector of i-th of sample, dimension n, then the expression formula of decoder are as follows:
In formula: W ' is the weight matrix of hidden layer and output layer；B ' is the threshold matrix of hidden layer and output layer；Sg is solution
The neuron activation functions of code device.
Autocoder reaches feature extraction by minimizing the reconstructed error between output vector and input vector
The formula of purpose, reconstructed error is as follows:
Network weight and threshold value are constantly adjusted using gradient descent algorithm, reduces reconstructed error, formula is as follows:
In formula: l is learning rate；It indicatesLocal derviation is asked to weight W；It indicatesTo threshold
Value B seeks local derviation.
SAE is a kind of deep learning neural network model being made of AE stack stacking, and the output of lower layer AE will be used as upper
The input of layer AE.Gradually being abstracted for feature is realized by the stacking of AE, more compact, useful feature is ultimately formed, such as Fig. 9 institute
It is shown as stacking self-encoding encoder framework.WnFor the weight matrix of (n-1)th layer of hidden layer and n-th layer hidden layer, BnIt is implicit for (n-1)th layer
The threshold matrix of layer and n-th layer hidden layer.Training process is divided into greedy successively unsupervised pre-training and has two steps of supervision fine tuning
Suddenly.Greedy successively unsupervised pre-training obtains the initialization weight and threshold value of network by successively training, and the input of bottom AE is
Initial data, input data of the hidden layer output data as upper layer AE.
After the completion of being layered pre-training, hidden layer is stacked, input data is expressed as with output data relationship:
In formula: f is activation primitive, xiFor the input variable of i-th of training sample, W, B are respectively that layer-by-layer pre-training obtains
Netinit weight and threshold value,For the predicted value of i-th of sample.The error for constructing actual value and predicted value loses letter
Number, formula are as follows:
In formula: N is sample total number, yiFor the actual value of i-th of sample.By top-down backpropagation to entire
Network weight and threshold value are finely adjusted, and reduce the error of predicted value and actual value.
SAE can effectively extract the high-order feature of data by successively greedy unsupervised pre-training, preferably approach complexity
Function, and parameter optimization space is reduced, network parameter can be quickly obtained, the further feature learning ability of neural network is improved.
Compared to shallow-layer machine learning algorithm, deep layer network is when handling high dimensional data, because its complicated network structure is more
Add and be easy to produce overfitting problem, thus the generalization ability of limited model.Dropout is a kind of anti-over-fitting technology of mainstream,
Its basic thought are as follows: randomly choose a part of node in model training and do not work, these nodes will save last iteration
Weight, and output is set to 0.The weight that these nodes selected retain can restore during next iteration again before,
Random selection part of nodes repeats this process again.All certain variation will occur for network structure in each iterative process, adopt
With Dropout technology, randomly selects partial nerve member and do not work temporarily, as shown in Figure 10, reduce being total between specific node
Same-action alleviates dependence of the network output to specific node state, to prevent over-fitting.
In the small parameter perturbations stage, using top-down small lot RMSProp optimization method to the weight and threshold value of network
It is finely adjusted, until the number of iterations reaches setting value, the specific steps of which are as follows:
Step 1, global learning rate l, rate of decay ρ are set, cumulative variations r1=0, r2=0 are initialized.
Step 2, the small lot data set comprising a sample of M ' is chosen from training set, according to error loss function, is calculated
Gradient:
Step 3, accumulative squared gradient is calculated, as shown in formula (9):
In formula: ⊙ is by element product symbol.
Step 4, weight and threshold parameter are updated respectively:
Step 5, when the number of iterations reaches requirement, stop operation, otherwise return to step 2 and continue to execute calculating.
Processing medium temperature prediction control algolithm of the invention can be become with Efficient Characterization processing medium temperature deviation and deviation
Complicated function between rate, the cooling feed temperature of industrial cycle and cooling feed pressure setting value, fast and accurately to industry
Production scene processing medium temperature carries out PREDICTIVE CONTROL, while the algorithm has good generalization ability, becomes to different temperatures
The processing medium prediction for changing feature is adaptable.Compared with artificial experience is calculated and adjusted, processing medium temperature is greatly improved
The accuracy of control saves calculating time overhead, facilitates industrial site administrative staff and is grasped in each heat exchanger in real time
Processing medium temperature changing trend；Processing medium temperature prediction control algolithm of the invention includes based on storehouse autocoder
Processing medium temperature prediction model off-line training and application on site provide recirculating cooling water system cooling and give back to hydraulic pressure in heat exchanger
Poor setting value: storehouse autocoder belongs to deep learning neural network, and compared with traditional shallow-layer neural network, storehouse is automatic
Encoder efficiently solves a series of problems caused by traditional neural network stochastic parameter initializes, and can effectively excavate each number
According to implication relation, industrial site processing medium temperature prediction precise control is greatly improved；Compared to conventional stack
Autocoder method has more using the improvement storehouse autocoder algorithm that RMSProp optimization method carries out small parameter perturbations
Industrial site processing medium temperature prediction control precision can be improved in good generalization ability.It is raw that this method is applied to industry
Live recirculating cooling water system Energy Saving Control is produced, cooling can be quick and precisely provided and give back to differential water pressures setting value, be conducive to industry
Production link personnel grasp processing medium temperature changing trend in each heat exchanger in real time, it is most important that reduce circulating cooling
Water system energy consumption.
It should be pointed out that the above description is not a limitation of the present invention, the present invention is also not limited to the example above,
Variation, modification, addition or the replacement that those skilled in the art are made within the essential scope of the present invention, are also answered
It belongs to the scope of protection of the present invention.
Claims (10)
1. a kind of recirculated cooling water minimum differntial pressure energy-saving control system based on processing medium multi-temperature target, including circulating cooling
Water system and control energy conserving system；
The recirculating cooling water system includes along N number of cooling tower of circulation waterway setting, priming reservoir, water supply pump assembly, water outlet
A heat exchanger of pipe group, M ' and return pipe group；Cooling bay is provided in cooling tower；
The outlet pipe group includes N root priming reservoir water inlet pipe, L root priming reservoir outlet pipe, water main, M ' root subfeeder；
The return pipe group includes M ' root return branch, return main, tower return pipe on N root；
Any cooling tower correspondence is connect through a priming reservoir water inlet pipe with the priming reservoir, and the priming reservoir passes through side by side
Priming reservoir outlet pipe described in L root is connect with the water main, and the water main is through M ' root subfeeder to a heat exchanger of M '
It corresponds and supplies water；
Feed pump has been arranged in parallel on the priming reservoir outlet pipe；
Any heat exchanger correspondence is connect through a return branch with the return main, and the return main is through N arranged side by side
Tower return pipe connects one to one with N number of cooling tower on root；
It is provided with processing medium temperature sensor in a heat exchanger of M ', the processing medium temperature sensor is each for obtaining
Kind processing medium real time temperature detected value；
Cooling feed temperature sensor and cooling feed pressure sensor, the cooling water supply are provided on the water main
For temperature sensor for obtaining cooling feed temperature detected value, the cooling feed pressure sensor is cooling to hydraulic pressure for obtaining
Power detected value；
Cooling backwater temperature sensor and cooling backwater pressure sensor, the cooling backwater are provided on the return main
Temperature sensor cools back hydraulic pressure for obtaining for obtaining cooling backwater temperature detection value, the cooling backwater pressure sensor
Power detected value；
A upper tower valve is respectively arranged on the upper tower return pipe described in N root；
It is characterized by: including processing medium temperature least favorable point selector, processing medium temperature in the control energy conserving system
Controller and give back to differential water pressures inner ring PID controller；The processing medium temperature least favorable point selector is cold according to the circulation
But the collected processing medium temperature deviation value sequence of water system obtains processing medium temperature least favorable point；
The processing medium temperature controller inputs work according to the processing medium temperature least favorable point, in conjunction with field operational data
Skill medium temperature controller obtains cooling and gives back to differential water pressures setting value；
After the cooling feed pressure detected value and cooling backwater pressure detection value that the recirculating cooling water system detects make difference
It obtains cooling and gives back to differential water pressures detected value；The cooling gives back to differential water pressures detected value and cooling give back to after differential water pressures setting value makees difference obtains
To cooling give back to differential water pressures deviation；It is described give back to differential water pressures inner ring PID controller according to cooling give back to differential water pressures deviation
The aperture of all upper tower valves is adjusted, to change the outlet stream amount of water supply pump assembly.
2. the recirculated cooling water minimum differntial pressure Energy Saving Control system according to claim 1 based on processing medium multi-temperature target
System, it is characterised in that: the processing medium temperature deviation value sequence includes a processing medium temperature deviation value of M ', any work
Skill medium temperature deviation is equal to the corresponding processing medium real time temperature detected value and sets with the corresponding processing medium temperature
The difference of value；
The processing medium temperature deviation change rate sequence be include a processing medium temperature deviation change rate of M ', wherein
Any processing medium temperature deviation change rate be the corresponding two neighboring detection period temperature change value of processing medium and
The ratio of a upper detection period.
3. a kind of recirculated cooling water minimum differntial pressure energy conservation control according to claim 1 based on processing medium multi-temperature target
The control method of system processed, it is characterised in that specific steps are as follows:
S1: the setting sampling period carries out field operational data acquisition to the recirculating cooling water system；
The recirculating cooling water system passes through the M ' in a heat exchanger of processing medium temperature sensor collection in worksite M '
A processing medium real time temperature detected value；
The recirculating cooling water system passes through the cooling feed temperature detected value of the cooling feed temperature sensor collection in worksite；
The recirculating cooling water system passes through the cooling feed pressure detected value of the cooling feed pressure sensor collection in worksite；
The recirculating cooling water system passes through the cooling backwater temperature sensor collection in worksite cooling backwater temperature detection value；
The recirculating cooling water system passes through the cooling backwater pressure sensor collection in worksite cooling backwater pressure detection value；
S2: according to a processing medium real time temperature detected value of the M ' of step S1 collection in worksite and corresponding a processing medium temperature of M '
Setting value is spent, a processing medium temperature deviation value of M ' and corresponding a processing medium temperature deviation change rate of M ' are obtained；
A processing medium temperature deviation value of M ' constitutes the processing medium temperature deviation value sequence；
A processing medium temperature deviation rate of change value of M ' constitutes the processing medium temperature deviation change rate sequence；
S3: the temperature deviation in the processing medium temperature least favorable point selector selection processing medium temperature deviation value sequence is most
Small value is used as processing medium temperature least favorable point, and it is inclined to obtain the corresponding processing medium temperature of processing medium temperature least favorable point
Poor change rate and corresponding heat exchanger；S4: processing medium temperature controller goes to obtain according to the processing medium temperature least favorable point
Processing medium temperature deviation value, the processing medium temperature deviation change rate for taking corresponding heat exchanger, by the scene of corresponding heat exchanger
Operation data inputs the corresponding processing medium temperature prediction model being embedded in processing medium temperature controller, obtain it is cooling to
Return water pressure differential resetting value and processing medium temperature prediction value；
S5: the cooling feed pressure detected value and cooling backwater pressure detection value are obtained into the cooling differential water pressures that give back to as difference and detected
Value；The cooling gives back to differential water pressures setting value and the cooling differential water pressures detected value that gives back to is made to obtain cooling after difference to give back to differential water pressures deviation
Value, the cooling give back to and give back to differential water pressures inner ring PID controller described in differential water pressures deviation feeding, the aperture of tower valve in adjusting, from
And change the outlet stream amount of water supply pump assembly.
4. the recirculated cooling water minimum differntial pressure Energy Saving Control system according to claim 3 based on processing medium multi-temperature target
The control method of system, it is characterised in that find out processing medium temperature using processing medium temperature least favorable point selector in step S3
Specific steps of the minimum value as processing medium temperature least favorable point in deviation value sequence are as follows:
S31: initialization amounts to a processing medium temperature of M ' if a processing medium temperature deviation value of M ' forms a difference group
Deviation enables Wk=M '；K=1；
S32: enabling Wk+1=Wk+X, divides exactly Wk+1 by M ', and X is the wireless big vacancy of difference of filling；And X be equal to 0~
M'-1；
S33: W is calculatedk+2=Wk+1/M'；
S34: from Wk+2In group, using comparison method is intersected, minimum is found out from a processing medium temperature deviation value of each group of M '
Value, obtains Wk+2A processing medium temperature deviation value；
S35: judge Wk+2Whether 1 is equal to；If so, using the processing medium temperature deviation value as processing medium temperature least favorable point；
Otherwise, k=k+2 is enabled；Return step S32.
5. the recirculated cooling water minimum differntial pressure Energy Saving Control system according to claim 3 based on processing medium multi-temperature target
The control method of system, it is characterised in that a processing medium temperature prediction model deep learning neural network of M ' in step S4 is built
Vertical step are as follows:
S411: the historical data that recirculating cooling water system runs the historical data of generation within X sampling period, and will acquire,
Training data after being screened according to screening conditions, as processing medium temperature prediction model in heat exchanger；
S412: from historical data according to determine characteristic variable, by the historical data of heat exchanger processing medium temperature deviation,
Processing medium temperature deviation change rate, cooling feed temperature detected value, cooling give back to differential water pressures detected value as input data, and
Normalizing data set is obtained after carrying out data normalization processing, and the normalizing data set is divided into training sample set and test sample
Collection；
S413: based on autocoder is stacked, successively greedy unsupervised pre-training is carried out to training sample set, obtains deep learning
The input layer of neural network initialization and weight matrix W, input layer and the hidden layer threshold value matrix B of hidden layer；
S414: small parameter perturbations are carried out: the weight matrix W, defeated of input layer and hidden layer to the initialization of deep learning neural network
Enter layer to be finely adjusted with hidden layer threshold value matrix B, until the number of iterations reaches the number of iterations maximum value, obtain based on stacking
The initial process medium temperature prediction model of autocoder；
S415: assessing the initial process medium temperature prediction model that step S414 is obtained using test sample data set,
It obtains based on the processing medium temperature prediction model for stacking autocoder.
6. the recirculated cooling water minimum differntial pressure Energy Saving Control system according to claim 5 based on processing medium multi-temperature target
The control method of system, it is characterised in that the characteristic variable include processing medium temperature deviation value in any one heat exchanger,
Cooling in processing medium temperature deviation change rate, processing medium real time temperature detected value and recirculating cooling water system is to water temperature
Degree detected value, cooling give back to differential water pressures detected value.
7. the recirculated cooling water minimum differntial pressure Energy Saving Control system according to claim 6 based on processing medium multi-temperature target
The control method of system, it is characterised in that the screening conditions are that processing medium temperature is in safe and energy-efficient temperature value section
Historical data, described safe and energy-efficient temperature value section is in processing medium temperature threshold section.
8. the recirculated cooling water minimum differntial pressure Energy Saving Control system according to claim 5 based on processing medium multi-temperature target
The control method of system, it is characterised in that in step S413, when carrying out successively greedy unsupervised pre-training to training sample set, will instruct
Practice sample set and be divided into P group small lot training sample, be successively trained, and use Dropout technology, randomly selects partial nerve
First break-off, successively iteration, successively training obtain the input layer of deep learning neural network initialization and the weight of hidden layer
Matrix W, input layer and hidden layer threshold value matrix B.
9. the recirculated cooling water minimum differntial pressure Energy Saving Control system according to claim 5 based on processing medium multi-temperature target
The control method of system, it is characterised in that in step S414, when small parameter perturbations, optimized using top-down small lot RMSPROP
Weight matrix W, input layer and the hidden layer threshold value square of input layer and hidden layer that method initializes deep learning neural network
Battle array B is finely adjusted.
10. the recirculated cooling water minimum differntial pressure Energy Saving Control according to claim 3 based on processing medium multi-temperature target
The control method of system, it is characterised in that give back to differential water pressures inner ring PID controller described in step S5 and the upper tower valve is opened
The specific control content of the outlet stream amount of degree and the water supply pump assembly are as follows:
If the cooling differential water pressures deviation value P that gives back to, greater than 0, the upper tower valve of differential water pressures inner ring PID controller sending that gives back to tunes up control
Signal processed increases valve opening for controlling upper tower valve；
If the cooling differential water pressures deviation value P that gives back to, equal to 0, the differential water pressures inner ring PID controller that gives back to keeps former state of a control；
If cooling give back to differential water pressures deviation value P less than 0, the upper tower valve of differential water pressures inner ring PID controller sending that gives back to turns control down
Signal processed reduces valve opening for controlling upper tower valve.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910251616.9A CN109917656B (en) | 2019-03-29 | 2019-03-29 | Circulating cooling water minimum pressure difference energy-saving control system and method based on process medium multi-temperature target |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910251616.9A CN109917656B (en) | 2019-03-29 | 2019-03-29 | Circulating cooling water minimum pressure difference energy-saving control system and method based on process medium multi-temperature target |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109917656A true CN109917656A (en) | 2019-06-21 |
| CN109917656B CN109917656B (en) | 2022-03-01 |
Family
ID=66967827
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910251616.9A Expired - Fee Related CN109917656B (en) | 2019-03-29 | 2019-03-29 | Circulating cooling water minimum pressure difference energy-saving control system and method based on process medium multi-temperature target |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109917656B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112434298A (en) * | 2021-01-26 | 2021-03-02 | 浙江大学 | Network threat detection system based on self-encoder integration |
| CN116301117A (en) * | 2023-05-24 | 2023-06-23 | 荣耀终端有限公司 | Temperature control method and electronic equipment |
| WO2023138325A1 (en) * | 2022-01-18 | 2023-07-27 | 光大环保技术装备（常州）有限公司 | Cooling medium temperature regulation and control method, system and device for circulating cooling system |
| CN117234071A (en) * | 2023-11-16 | 2023-12-15 | 中控技术股份有限公司 | Circulating water system energy-saving optimization method and device based on temperature and pressure control |
Citations (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5691896A (en) * | 1995-08-15 | 1997-11-25 | Rosemount, Inc. | Field based process control system with auto-tuning |
| US5701554A (en) * | 1994-06-10 | 1997-12-23 | Seiko Epson Corporation | Fixing apparatus having controller for setting a target temperature and for estimating the amount of heat transferred to a pressure roller |
| JP2002147914A (en) * | 2000-11-13 | 2002-05-22 | Sanyo Engineering Kk | Low temperature cool-storage medium |
| US6619066B1 (en) * | 1999-02-24 | 2003-09-16 | Hachiyo Engineering Co., Ltd. | Heat pump system of combination of ammonia cycle carbon dioxide cycle |
| US20080046111A1 (en) * | 2006-08-15 | 2008-02-21 | Yuki Kataoka | Heat processing apparatus, heat processing method, computer program and storage medium |
| JP2008232467A (en) * | 2007-03-16 | 2008-10-02 | Toshiba Corp | Air-conditioning control system |
| CN101382773A (en) * | 2008-10-17 | 2009-03-11 | 东华大学 | Anti-time lag flase-proportional integral control system and method |
| CN101893296A (en) * | 2009-05-22 | 2010-11-24 | 富士电机系统株式会社 | Accurate temperature adjusting system and control device thereof |
| CN102147146A (en) * | 2011-04-22 | 2011-08-10 | 黄真银 | Digital integrated intelligent control system of central air conditioner and adjusting method thereof |
| CN202133298U (en) * | 2011-05-03 | 2012-02-01 | 戴军 | Energy-saving control unit for condenser system of electric power plant |
| WO2012049702A1 (en) * | 2010-10-12 | 2012-04-19 | 三菱電機株式会社 | Air conditioner |
| US20120143528A1 (en) * | 2007-07-30 | 2012-06-07 | Lawrence Kates | Portable method and apparatus for monitoring refrigerant-cycle systems |
| CN102980272A (en) * | 2012-12-08 | 2013-03-20 | 珠海派诺科技股份有限公司 | Air conditioner system energy saving optimization method based on load prediction |
| CN104566765A (en) * | 2013-10-16 | 2015-04-29 | 嘉日国际集团控股有限公司 | Overall energy saving control method of central air conditioner |
| WO2015136172A1 (en) * | 2014-03-13 | 2015-09-17 | Peugeot Citroen Automobiles Sa | Method for predicting the thermal status of a starter |
| CN205242674U (en) * | 2015-11-26 | 2016-05-18 | 天津大学 | Envelope adjustable laboratory that loads |
| CN106152250A (en) * | 2016-08-16 | 2016-11-23 | 河南柴油机重工有限责任公司 | Blood circulation that a kind of heat exchange station automatically controls and circulation controlling means |
| CN106440549A (en) * | 2016-05-30 | 2017-02-22 | 靳凯 | Control system for increasing comprehensive energy efficiency ratio of central air conditioning system |
| CN107192097A (en) * | 2017-06-06 | 2017-09-22 | 珠海格力电器股份有限公司 | The control method and device of air conditioning water system |
| CN107976088A (en) * | 2016-10-24 | 2018-05-01 | 李宁 | A kind of closed cycle cooling process and device |
| JP6324560B1 (en) * | 2017-02-22 | 2018-05-16 | クラフトワーク株式会社 | Ice making apparatus and ice production method |
| CN108507245A (en) * | 2018-04-04 | 2018-09-07 | 重庆大学 | A kind of the full-automatic circulating cooling control system and control method of nesting closed loop |
| CN108643278A (en) * | 2018-06-22 | 2018-10-12 | 安徽机电职业技术学院 | A kind of energy-saving Township water supply system and its water-supply control |
| CN108990383A (en) * | 2018-08-15 | 2018-12-11 | 北京建筑大学 | A kind of data center's air-conditioning system forecast Control Algorithm |
| CN109460861A (en) * | 2018-10-18 | 2019-03-12 | 扬州大学 | A kind of chemical circulation cooling water system water pump fan combination frequency conversion optimization operating scheme determines method |
-
2019
- 2019-03-29 CN CN201910251616.9A patent/CN109917656B/en not_active Expired - Fee Related
Patent Citations (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5701554A (en) * | 1994-06-10 | 1997-12-23 | Seiko Epson Corporation | Fixing apparatus having controller for setting a target temperature and for estimating the amount of heat transferred to a pressure roller |
| US5691896A (en) * | 1995-08-15 | 1997-11-25 | Rosemount, Inc. | Field based process control system with auto-tuning |
| US6619066B1 (en) * | 1999-02-24 | 2003-09-16 | Hachiyo Engineering Co., Ltd. | Heat pump system of combination of ammonia cycle carbon dioxide cycle |
| JP2002147914A (en) * | 2000-11-13 | 2002-05-22 | Sanyo Engineering Kk | Low temperature cool-storage medium |
| US20080046111A1 (en) * | 2006-08-15 | 2008-02-21 | Yuki Kataoka | Heat processing apparatus, heat processing method, computer program and storage medium |
| JP2008232467A (en) * | 2007-03-16 | 2008-10-02 | Toshiba Corp | Air-conditioning control system |
| US20120143528A1 (en) * | 2007-07-30 | 2012-06-07 | Lawrence Kates | Portable method and apparatus for monitoring refrigerant-cycle systems |
| CN101382773A (en) * | 2008-10-17 | 2009-03-11 | 东华大学 | Anti-time lag flase-proportional integral control system and method |
| CN101893296A (en) * | 2009-05-22 | 2010-11-24 | 富士电机系统株式会社 | Accurate temperature adjusting system and control device thereof |
| WO2012049702A1 (en) * | 2010-10-12 | 2012-04-19 | 三菱電機株式会社 | Air conditioner |
| CN102147146A (en) * | 2011-04-22 | 2011-08-10 | 黄真银 | Digital integrated intelligent control system of central air conditioner and adjusting method thereof |
| CN202133298U (en) * | 2011-05-03 | 2012-02-01 | 戴军 | Energy-saving control unit for condenser system of electric power plant |
| CN102980272A (en) * | 2012-12-08 | 2013-03-20 | 珠海派诺科技股份有限公司 | Air conditioner system energy saving optimization method based on load prediction |
| CN104566765A (en) * | 2013-10-16 | 2015-04-29 | 嘉日国际集团控股有限公司 | Overall energy saving control method of central air conditioner |
| WO2015136172A1 (en) * | 2014-03-13 | 2015-09-17 | Peugeot Citroen Automobiles Sa | Method for predicting the thermal status of a starter |
| CN205242674U (en) * | 2015-11-26 | 2016-05-18 | 天津大学 | Envelope adjustable laboratory that loads |
| CN106440549A (en) * | 2016-05-30 | 2017-02-22 | 靳凯 | Control system for increasing comprehensive energy efficiency ratio of central air conditioning system |
| CN106152250A (en) * | 2016-08-16 | 2016-11-23 | 河南柴油机重工有限责任公司 | Blood circulation that a kind of heat exchange station automatically controls and circulation controlling means |
| CN107976088A (en) * | 2016-10-24 | 2018-05-01 | 李宁 | A kind of closed cycle cooling process and device |
| JP6324560B1 (en) * | 2017-02-22 | 2018-05-16 | クラフトワーク株式会社 | Ice making apparatus and ice production method |
| CN107192097A (en) * | 2017-06-06 | 2017-09-22 | 珠海格力电器股份有限公司 | The control method and device of air conditioning water system |
| CN108507245A (en) * | 2018-04-04 | 2018-09-07 | 重庆大学 | A kind of the full-automatic circulating cooling control system and control method of nesting closed loop |
| CN108643278A (en) * | 2018-06-22 | 2018-10-12 | 安徽机电职业技术学院 | A kind of energy-saving Township water supply system and its water-supply control |
| CN108990383A (en) * | 2018-08-15 | 2018-12-11 | 北京建筑大学 | A kind of data center's air-conditioning system forecast Control Algorithm |
| CN109460861A (en) * | 2018-10-18 | 2019-03-12 | 扬州大学 | A kind of chemical circulation cooling water system water pump fan combination frequency conversion optimization operating scheme determines method |
Non-Patent Citations (6)
| Title |
|---|
| CHUN-LIEN SU，等: "Evaluation_of_Differential_Pressure_Setpoint_of_Chilled_Water_Pumps_in_Clean_Room_HVAC_Systems_for_Energy_Savings_in_High-Tech_Industries", 《IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS》 * |
| HONGLI LIU，等: "Design of LT-MED Seawater Desalination Temperature Control System Based on Dynamic Matrix Predictive Fuzzy PID Control Algorithm", 《PROCEEDINGS OF 2018 IEEE INTERNATIONAL CONFERENCE ON MECHATRONICS AND AUTOMATION》 * |
| MUHAMMAD NAZAR UL ISLAM，等: "The_effects_of_temperature_gradient_on_transient_behavior_of_low-range_differential_air_pressure_calibration_setup", 《2015 IEEE INTERNATIONAL CONFERENCE ON INDUSTRIAL TECHNOLOGY (ICIT)》 * |
| 唐鑫，等: "中央空调房间温度智能PID控制的仿真研究", 《计算机仿真》 * |
| 孔令坤: "基于AM3517的换热站节能控制系统研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 * |
| 曹华，等: "中央空调冷却水、冷冻水系统的变频节能分析", 《变频调速与应用》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112434298A (en) * | 2021-01-26 | 2021-03-02 | 浙江大学 | Network threat detection system based on self-encoder integration |
| WO2023138325A1 (en) * | 2022-01-18 | 2023-07-27 | 光大环保技术装备（常州）有限公司 | Cooling medium temperature regulation and control method, system and device for circulating cooling system |
| CN116301117A (en) * | 2023-05-24 | 2023-06-23 | 荣耀终端有限公司 | Temperature control method and electronic equipment |
| CN116301117B (en) * | 2023-05-24 | 2023-10-20 | 荣耀终端有限公司 | Temperature control method and electronic equipment |
| CN117234071A (en) * | 2023-11-16 | 2023-12-15 | 中控技术股份有限公司 | Circulating water system energy-saving optimization method and device based on temperature and pressure control |
| CN117234071B (en) * | 2023-11-16 | 2024-01-26 | 中控技术股份有限公司 | Circulating water system energy-saving optimization method and device based on temperature and pressure control |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109917656B (en) | 2022-03-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109917656A (en) | Recirculated cooling water minimum differntial pressure energy-saving control system and method based on processing medium multi-temperature target | |
| CN111256294B (en) | Model prediction-based optimization control method for combined operation of water chilling unit | |
| CN110288164B (en) | Predictive control method for building air-conditioning refrigeration station system | |
| CN108990383B (en) | Predictive control method for air conditioning system of data center | |
| CN105910169B (en) | District heating system regulating of heating net method and system based on mechanism model PREDICTIVE CONTROL | |
| Yabanova et al. | Development of ANN model for geothermal district heating system and a novel PID-based control strategy | |
| CN102508474B (en) | Circulated cooling water operation optimization control system for industrial enterprise | |
| CN113266869B (en) | Real-time optimization regulation and control method of thermoelectric combined heating system based on digital twinning technology | |
| CN110392515A (en) | A kind of Cooling and Heat Source equipment room energy-conserving control method and system based on historical data | |
| CN110008575A (en) | Recirculating cooling water system processing medium multi-temperature target set point switches multi-parameter prediction control algolithm | |
| CN103411473B (en) | Industrial circulating water system combination energy-saving method and industrial circulating water combination energy-saving system | |
| CN105864016B (en) | A kind of more water pump distributing system variable water volume runing adjustment methods of open type | |
| CN102102962A (en) | Energy-saving regulating and control method and device of industrial cooling water towering pump | |
| CN108563877B (en) | Integral modeling and optimal control integrated method for solar lithium bromide refrigerating unit | |
| CN113822496A (en) | Multi-unit thermal power plant heat supply mode and parameter online optimization method | |
| CN110454373A (en) | A kind of system of sea cooling water variable frequency pump optimizing operation method | |
| CN105759773A (en) | Petrochemical enterprise factory model extension method based on object oriented analysis | |
| CN111767677A (en) | GA algorithm-based cascade pump station group lift optimal distribution method | |
| CN113152595B (en) | Variable-frequency constant-pressure water supply system and energy-saving control method thereof | |
| CN113218040A (en) | Energy efficiency improvement control method for central air-conditioning system | |
| CN115903712A (en) | Energy-saving optimization method and optimization control system suitable for industrial refrigeration system | |
| CN116499023A (en) | Intelligent control method and system for geothermal coupling solar heating station | |
| CN217813611U (en) | System for rapidly improving load response rate of thermal power generating unit based on heat storage tank | |
| CN115577828A (en) | Air conditioner refrigerating station system group control method based on data-driven modeling and optimization | |
| CN115423191A (en) | Room temperature model prediction control method based on neural network and constrained by room heat load |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220301 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |