Central heating system heat supply network multivariable optimization control method based on LSTM model

Abstract

The invention belongs to the technical field of heat supply system heat supply network intelligent control, and discloses a centralized heat supply system heat supply network multivariable optimization control method based on an LSTM model, and the method comprises the steps: 1, collecting parameter historical operation data, carrying out the preprocessing of the data, and obtaining a standardized data set suitable for a recurrent neural network; 2, dividing the standardized data set into a training set, a verification set and a test set, and modeling the data set based on an LSTM model; 3, establishing a judgment condition of the steady state of the heat supply network, and calculating to obtain a stable temperature value of the secondary side return water of the heat exchange station after the parameters of the executing mechanism of the heat exchange station are changed in combination with the prediction model of the secondary side return water temperature of the heat exchange station; and 4, performing iterative optimization calculation on the parameters of the execution mechanism based on the temperature prediction model, and finally outputting the optimization control parameters of the execution mechanism. The method is of great significance in improving the automation and intelligence level of the heat supply system and improving the stability and economy of the heat supply system.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent control of a heating network of a heating system, and in particular relates to a multivariable optimization control method of a heating network of a centralized heating system based on an LSTM model. Background Art

[0002] The heating network of the heating system includes several heat exchange stations, and the working conditions and load conditions of each heat exchange station are different. The current heating system only implements single-loop control for a single heat exchange station, and the control systems of each heat exchange station are independent of each other and have no correlation, ignoring the correlation within the heating network, the uncertainty of the model and some nonlinearities, resulting in control oscillation and excessive dead zone between the heat exchange stations in actual applications.

[0003] These phenomena can only be solved by manual observation and manual adjustment of the valve opening and pump frequency setting values ​​of different heat exchange stations to solve the hydraulic imbalance of the pipeline network. This method has a long adjustment cycle and cannot flexibly respond to the complex changes in the heating system. In view of this, it is urgent to develop a multivariable process control system that needs to be optimized to better meet the joint control requirements of multiple heat exchange stations.

[0004] In actual work, the secondary return water temperature of the heat exchange station is generally used to evaluate the heating quality. Among similar prediction technologies, some methods do not consider the delay of system temperature changes, while some methods focus more on the impact of the load side on system changes. The output results have limited application value in guiding the regulation of heat exchange stations. Therefore, establishing a system model that considers the impact of temperature delay on the secondary return water temperature is also a problem that the invention intends to solve.

[0005] In summary, the joint control method of multiple heat exchange station actuators considering temperature delay is a necessary and sufficient condition for achieving high-quality regulation of the heating system. With the development of communication technology and artificial intelligence, it is possible to simultaneously obtain the operating parameters of multiple heat exchange stations, and quickly and accurately implement a multivariable process control system based on the correlation between the parameters of multiple heat exchange station actuators and the secondary return water temperature through AI algorithms. Summary of the invention

[0006] In order to solve the problems in the related art, the present application provides a multivariable optimization control method for the heating network of a centralized heating system based on an LSTM model, which solves the defects mentioned in the background technology.

[0007] The technical solution is as follows:

[0008] The invention discloses a multivariable optimization control method for a heat network of a centralized heating system based on an LSTM model, and the steps include: step 1, collecting historical operation data of ambient temperature, main network supply / return water temperature, main network water pump frequency, primary side supply / return water temperature of a heat exchange station, secondary side supply / return water temperature of a heat exchange station, and parameters of an actuator of a heat exchange station, preprocessing the data, and obtaining a standardized data set suitable for a recurrent neural network; step 2, dividing the standardized data set into a training set, a validation set, and a test set, modeling the data set based on an LSTM model, and obtaining a prediction model for the secondary side return water temperature of a heat exchange station; step 3, constructing a judgment condition for a steady state of a heat network, and combining the above-mentioned prediction model for the secondary side return water temperature of a heat exchange station to calculate the stable temperature value of the secondary side return water of the heat exchange station after the parameters of the actuator of the heat exchange station are changed; step 4, according to the deviation between the secondary side return water temperature of the heat exchange station and the set value, based on the temperature prediction model, iteratively optimizing and calculating the parameters of the actuator, until the deviation between the predicted stable temperature of the secondary side return water and the set value of the secondary side return water is less than a threshold, and finally outputting the optimized control parameters of the actuator.

[0009] For further improvement, the parameter types and quantities of the data set in step one include: 1 ambient temperature, 1 main network water supply temperature, 1 main network return water temperature, k main network water pump frequencies, m heat exchange station primary side supply water temperatures, m heat exchange station primary side return water temperatures, m heat exchange station secondary side supply water temperatures, m heat exchange station secondary side return water temperatures, and m heat exchange station actuator setting parameters. The data dimension is 3+k+5m, denoted as n.

[0010] For further improvement, the data set of step 1 is the set of the above parameters in different time series, and t is used to represent a certain moment in the time series, that is, the input feature vector of the LSTM model is recorded as Where n represents the data dimension and t represents the time.

[0011] For further improvement, in step 2, for the original operation data set, find any sample in which the setting parameters of the heat exchange station actuator change, select the sample and the sample data of N moments after it, that is, I t ,I t+1 ,...,I t+N , as the dataset used for LSTM modeling.

[0012] For further improvement, in step 2, according to the historical operation data of the central heating network, a LSTM-based heat exchange station secondary return water temperature prediction model is constructed; the input sample of the model is The output of the heat exchange station secondary side return water temperature prediction result is The vector dimension m is the number of heat exchange stations.

[0013] Further improvement, in step 3, the temperature stability judgment condition is that the difference between the secondary return water temperature of all heat exchange stations and the previous moment is lower than the threshold temperature, that is, If the predicted temperature value at time t+1 does not meet this condition, repeat step 3 to calculate the temperature value at time t+2; repeat the above process until the above condition is met at time t+s and the stable return water temperature value on the secondary side is output.

[0014] Further improvement, the process of applying the prediction model of the secondary return water temperature of the heat exchange station is described as follows: 1) Input the measured value of the secondary return water temperature of the heat exchange station at time t When it is equal to the secondary side temperature setting value of the heat exchange station (TR m ) is greater than or equal to the action threshold (CTS), that is, Adjust the preset parameter value of the actuator; 2) Use the secondary return water temperature measurement value and the adjusted actuator preset value as the input of the LSTM model to train the secondary return water temperature value at the time of temperature stability 3) Determine the stable value of the secondary return water temperature With the set value TR m If the deviation If the actuator setting parameters are output, otherwise the actuator preset values ​​will be adjusted continuously until the actuator setting parameters are met. After that, the actuator setting parameters are output.

[0015] By adopting the above technical solution, compared with the prior art, the beneficial effects of the technology of the present invention are:

[0016] The present invention adopts a neural network training model based on the LSTM layer and a temperature difference threshold judgment function to mine the sample data association relationship of two types of influencing factors that exist simultaneously, namely, time series and multivariable coupling relationships. After the outdoor temperature, heat source parameters of the heating system, primary side supply / return water temperature and parameter setting values ​​of different heat exchange station actuators change, the return water temperature of the secondary side after a certain time interval can be predicted in time. The temperature prediction model is repeatedly executed through the temperature difference threshold judgment function to obtain the actuator setting value combination when the secondary side return water temperature difference of all heat exchange stations meets the threshold requirement. The combination is output as the multivariable optimization control value of the heating network of the heating system to the control system of each heat exchange station, so as to quickly realize the system balance of the heat supply regulation of the heat exchange station, and achieve the effect of fast, intelligent and overall energy saving when multiple thermal power stations in the heating system are jointly regulated under the condition that the heating quality meets the standard.

[0017] The present invention uniformly considers the influence of outdoor environmental parameters, the changes in the setting parameters of several heat exchange station actuators and the delay of temperature changes on the secondary return water temperature of the heat exchange station, and constructs a heat exchange station secondary return water temperature prediction model based on the LSTM model. After multiple iterative calculations, the parameters of the heat exchange station actuators are calculated when the output temperature difference meets the threshold requirements. The present invention is of great significance for improving the automation and intelligence level of the heating system and improving the stability and economy of the heating system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0019] Figure 1 It is a structural principle block diagram of the present invention.

[0020] Figure 2 It is a flow chart of the present invention.

[0021] Figure 3 It is a schematic diagram of the principle of a multivariable optimization control method for a central heating system heat network based on LSTM return water temperature prediction in which the heat exchange station actuator is a valve in the present invention.

[0022] Figure 4 It is a schematic diagram of the principle of a multivariable optimization control method for a central heating system heat network based on LSTM return water temperature prediction in which the heat exchange station actuator in the present invention is a variable frequency water pump. DETAILED DESCRIPTION

[0023] The preferred embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the preferred embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0024] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0025] It should be understood that although the terms "first", "second", "third", etc. may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0026] like Figure 1 As shown, the multivariable optimization control method of the central heating system heat network based on the LSTM model includes the following steps:

[0027] Step 1: Collect historical operating data of ambient temperature, main network supply / return water temperature, main network water pump frequency, primary side supply / return water temperature of heat exchange station, secondary side supply / return water temperature of heat exchange station, and actuator parameters of heat exchange station, and preprocess the data to obtain a standardized data set suitable for recurrent neural network.

[0028] The parameter types and quantities of the data set in step 1 include: 1 ambient temperature, 1 main network water supply temperature, 1 main network return water temperature, k main network water pump frequencies, m heat exchange station primary side water supply temperature, m heat exchange station primary side return water temperature, m heat exchange station secondary side water supply temperature, m heat exchange station secondary side return water temperature, and m heat exchange station actuator setting parameters. The data dimension is 3+k+5m, denoted as n. The data set in step 1 is a collection of the above parameters in different time series. t is used to represent a certain moment in the time series, that is, the LSTM model input feature vector is denoted as Where n represents the data dimension and t represents the time.

[0029] Step 2: Divide the standardized data set into training set, validation set and test set, model the data set based on the LSTM model, and obtain the secondary side return water temperature prediction model of the heat exchange station.

[0030] In step 2, for the original operation data set, find any sample in which the setting parameters of the heat exchange station actuator change, select the sample and the sample data of N moments after it, that is, I t ,I t+1 ,...,I t+N , as the data set used for LSTM modeling. In step 2, based on the historical operation data of the central heating network, a LSTM-based heat exchange station secondary return water temperature prediction model is constructed; the input sample of the model is The output of the heat exchange station secondary side return water temperature prediction result is The vector dimension m is the number of heat exchange stations.

[0031] Step 3: Construct the judgment condition of the steady state of the heat network, combine the above-mentioned secondary side return water temperature prediction model of the heat exchange station, and calculate the stable temperature value of the secondary side return water of the heat exchange station after the parameters of the heat exchange station actuator are changed.

[0032] In step 3, the condition for temperature stability is that the difference between the secondary return water temperature of all heat exchange stations and the previous moment is lower than the threshold temperature, that is, If the predicted temperature value at time t+1 does not meet this condition, repeat step 3 to calculate the temperature value at time t+2; repeat the above process until the above condition is met at time t+s and the stable return water temperature value on the secondary side is output.

[0033] Step 4: According to the deviation between the secondary return water temperature of the heat exchange station and the set value, based on the temperature prediction model, the actuator parameters are iteratively optimized and calculated until the deviation between the predicted secondary return water stable temperature and the secondary return water set value is less than the threshold value, and finally the optimized control parameters of the actuator are output.

[0034] Example 1

[0035] This embodiment is a multivariable optimization control method for a heat network of a centralized heating system based on LSTM return water temperature prediction, in which the actuator of the heat exchange station is a valve.

[0036] like Figure 3 As shown in Figure 1, the central heating system includes 1 heat source, 2 main network variable frequency water pumps, 3 heat exchange stations, 3 heat exchangers, and 3 return water regulating valves. The dynamic environment system monitoring and collection data includes ambient temperature, main network supply / return water temperature, main network water pump frequency, heat exchange station primary side supply / return water temperature, heat exchange station secondary side supply / return water temperature, and heat exchange station return water regulating valve opening. Time-based data set vector I t The dimension is 20, denoted as Output result vector T t The dimension is 3, recorded as

[0037] When the opening degree of the return water regulating valve of any heat exchange station changes, the data sets of the above parameters in the dynamic ring system are obtained at the time of the change and 2 minutes, 4 minutes, 6 minutes, ..., 60 minutes after the change.

[0038] like Figure 2 As shown in the figure, a stable temperature prediction model flow chart is established, and the above filtered data sets are As input, the return water temperature prediction value at time t+1 is obtained by training based on the LSTM model

[0039] Stable temperature value calculation, judging whether the difference between the temperature value obtained after training and the temperature at the previous moment is less than the threshold value, the threshold value is 0.5℃, that is, when When Otherwise, repeat the previous step and calculate the return water temperature value at the next moment until the return water temperature at moment t+k satisfies After that, the output secondary side return water temperature All the above steps build a temperature prediction model based on the LSTM model.

[0040] Based on the temperature prediction model constructed in the above steps, the secondary return water temperature TSR3 of the heat exchange station is collected in real time. When the valve is opened, adjust the preset opening value VO3, use TSR3 and VO3 as the temperature prediction model input, and calculate the stable value of the secondary side return water temperature. If the return water temperature deviation is too large, When the temperature of the return water is lowered, the preset opening value of the heat exchange station valve is modified, and then iterative calculation is performed until the return water temperature deviation value is Finally, the valve opening setting value after multi-variable optimization of the heating network is output.

[0041] Example 2

[0042] The following is a specific implementation process of Example 2 of the present invention:

[0043] like Figure 4 As shown in Figure 1, the central heating system includes 2 heat sources, 2 main network variable frequency water pumps, 4 heat exchange stations, 4 heat exchangers, and 4 heat exchange station variable frequency water pumps. The dynamic environment system monitoring and collection data includes ambient temperature, main network supply / return water temperature, main network water pump frequency, heat exchange station primary side supply / return water temperature, heat exchange station secondary side supply / return water temperature, and heat exchange station primary side water pump frequency. Time-based data set vector I t The dimension is 25, denoted as Output result vector T t The dimension is 4, denoted as

[0044] When the frequency of the primary-side water pump of any heat exchange station changes, the data sets of the above parameters in the dynamic ring system are obtained at the time of the change and 2 minutes, 4 minutes, 6 minutes, ..., 60 minutes after the change.

[0045] like Figure 2 As shown in the figure, a stable temperature prediction model flow chart is established, and the above filtered data sets are As input, the return water temperature prediction value at time t+1 is obtained by training based on the LSTM model

[0046] Stable temperature value calculation, judging whether the difference between the temperature value obtained after training and the temperature at the previous moment is less than the threshold value, the threshold value is 0.4℃, that is, when When Otherwise, repeat the previous step and calculate the return water temperature value at the next moment until the return water temperature at moment t+k satisfies After that, the output secondary side return water temperature All the above steps build a temperature prediction model based on the LSTM model.

[0047] Based on the temperature prediction model constructed in the above steps, the secondary return water temperature TSR4 of the heat exchange station is collected in real time. When the preset frequency FPF4 of the variable frequency water pump of the heat exchange station is adjusted, TSR4 and FPF4 are used as the input of the temperature prediction model to calculate the stable value of the secondary side return water temperature. If the return water temperature deviation is too large, When the frequency of the heat exchange station variable frequency water pump is changed, the preset frequency value is modified, and then iterative calculation is performed until the return water temperature deviation value is Finally, the frequency setting value of the heat exchange station water pump is output after multivariable optimization of the heat network.

[0048] The present invention adopts a neural network training model based on the LSTM layer and a temperature difference threshold judgment function to mine the sample data association relationship of two types of influencing factors that exist simultaneously, namely, time series and multivariable coupling relationships. After the outdoor temperature, heat source parameters of the heating system, primary side supply / return water temperature and parameter setting values ​​of different heat exchange station actuators change, the return water temperature of the secondary side after a certain time interval can be predicted in time. The temperature prediction model is repeatedly executed through the temperature difference threshold judgment function to obtain the actuator setting value combination when the secondary side return water temperature difference of all heat exchange stations meets the threshold requirement. The combination is output as the multivariable optimization control value of the heating network of the heating system to the control system of each heat exchange station, so as to quickly realize the system balance of the heat supply regulation of the heat exchange station, and achieve the effect of fast, intelligent and overall energy saving when multiple thermal power stations in the heating system are jointly regulated under the condition that the heating quality meets the standard.

[0049] Those skilled in the art will readily conceive of other embodiments of the invention after considering the specification and practicing the invention herein. This application is intended to cover any variation, use, or adaptation of the invention that follows the general principles of the invention and includes common knowledge or customary techniques in the art not covered by the invention. The specification and examples are to be regarded as exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0050] It should be understood that the present invention is not limited to the precise structure described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A multivariable optimization control method for a central heating system heat network based on an LSTM model, characterized in that the steps include: Step 1: Collect historical operation data of ambient temperature, main network supply / return water temperature, main network water pump frequency, primary side supply / return water temperature of heat exchange station, secondary side supply / return water temperature of heat exchange station, and actuator parameters of heat exchange station, and pre-process the data to obtain a standardized data set suitable for recurrent neural network; Step 2: Divide the standardized data set into a training set, a validation set, and a test set, model the data set based on the LSTM model, and obtain a prediction model for the secondary side return water temperature of the heat exchange station; Step 3: construct the judgment condition of the steady state of the heat network, and combine the above-mentioned secondary return water temperature prediction model of the heat exchange station to calculate the stable temperature value of the secondary return water of the heat exchange station after the parameters of the heat exchange station actuator are changed; Step 4: According to the deviation between the secondary return water temperature of the heat exchange station and the set value, based on the temperature prediction model, the actuator parameters are iteratively optimized and calculated until the deviation between the predicted secondary return water stable temperature and the secondary return water set value is less than the threshold value, and finally the optimized control parameters of the actuator are output.

2. The multivariable optimization control method for the heating network of a central heating system based on the LSTM model according to claim 1 is characterized in that: The parameter types and quantities of the data set in step 1 include: 1 ambient temperature, 1 main network water supply temperature, 1 main network return water temperature, k main network water pump frequencies, m heat exchange station primary side water supply temperatures, m heat exchange station primary side return water temperatures, m heat exchange station secondary side water supply temperatures, m heat exchange station secondary side return water temperatures, and m heat exchange station actuator setting parameters. The data dimension is 3+k+5m, denoted as n.

3. The multivariable optimization control method for the heating network of a central heating system based on the LSTM model according to claim 2 is characterized in that: The data set of step 1 is the set of the above parameters in different time series. t is used to represent a certain moment in the time series, that is, the input feature vector of the LSTM model is recorded as Where n represents the data dimension and t represents the time.

4. The multivariable optimization control method for the heating network of a central heating system based on the LSTM model according to claim 3 is characterized in that: In step 2, for the original operation data set, find any sample in which the setting parameters of the heat exchange station actuator change, select the sample and the sample data of N moments after it, that is, I t ,I t+1 ,...,I t+N , as the dataset used for LSTM modeling.

5. The multivariable optimization control method for the heating network of a central heating system based on the LSTM model according to claim 4 is characterized in that: In step 2, according to the historical operation data of the central heating network, a LSTM-based heat exchange station secondary return water temperature prediction model is constructed; the input sample of the model is The output of the heat exchange station secondary side return water temperature prediction result is The vector dimension m is the number of heat exchange stations.

6. The multivariable optimization control method for the heating network of a central heating system based on the LSTM model according to claim 5 is characterized in that: In step 3, the condition for temperature stability is that the difference between the secondary return water temperature of all heat exchange stations and the previous moment is lower than the threshold temperature, that is, If the predicted temperature value at time t+1 does not meet the condition, repeat step 3 to calculate the temperature value at time t+2; Repeat the above process until the above conditions are met at time t+s and the stable return water temperature value on the secondary side is output.

7. The multivariable optimization control method for the heating network of a centralized heating system based on the LSTM model according to claim 6 is characterized in that: The application process of the secondary return water temperature prediction model of the heat exchange station is described as follows: 1) Input the measured value of the secondary return water temperature of the heat exchange station at time t When it is equal to the secondary side temperature setting value of the heat exchange station (TR m ) is greater than or equal to the action threshold (CTS), that is, Adjust the preset parameter values ​​of the actuator; 2) The secondary return water temperature measurement value and the adjusted actuator preset value are used as the input of the LSTM model to train the secondary return water temperature value at the time of temperature stability. 3) Determine the stable value of the secondary return water temperature With the set value TR m If the deviation If the actuator setting parameters are output, otherwise the actuator preset values ​​will be adjusted continuously until the actuator setting parameters are met. After that, the actuator setting parameters are output.

Images