Heat exchanger operation parameter optimization and dynamic response analysis method and device

Abstract

The embodiment of the invention relates to the field of heat exchangers, and provides a heat exchanger operation parameter optimization and dynamic response analysis method and device.The method comprises the steps that time series data of heat exchanger performance parameters are obtained, the time series data are preprocessed, and a dynamic response database is constructed; constructing a performance prediction model based on a long and short term memory network and a deep belief network, and training the performance prediction model by using a back propagation algorithm based on the dynamic response database; and the trained performance prediction model is combined with a multi-target particle swarm optimization algorithm, corresponding target operation parameters are reversely deduced for target output parameters of the heat exchanger, variable working condition dynamic response analysis and performance prediction of the heat exchanger are carried out, and optimal operation parameters are found. According to the embodiment of the invention, the precision and reliability of performance prediction of the heat exchanger are improved, the calculation cost is remarkably saved, the operation efficiency of the heat exchanger under various working conditions can be improved, and the application value is wide.

Description

Technical Field

[0001] The present disclosure relates to the technical field of heat exchangers, and in particular to a method and device for optimizing operating parameters and analyzing dynamic responses of heat exchangers. Background Art

[0002] Heat exchangers are key equipment in the fields of chemical industry, energy, HVAC, etc., and their performance is crucial to the overall efficiency of the system. However, the performance of heat exchangers under different working conditions is complex and changeable. Existing heat exchanger performance prediction methods are mainly based on empirical formulas and simplified models. Although the calculation is simple, the accuracy is low and it cannot accurately reflect the dynamic response characteristics of the heat exchanger in actual operation.

[0003] In addition, traditional transducer performance prediction methods often ignore the complex nonlinear relationship between operating parameters and heat exchanger performance, and are unable to perform the task of inferring the optimal operating parameter combination for the target output parameters. Summary of the invention

[0004] The present disclosure aims to solve at least one of the problems existing in the prior art, and to provide a reliable and accurate method and device for heat exchanger operating parameter optimization and dynamic response analysis, so as to solve the problem in the prior art that it is difficult to adjust the operating parameters according to the target operating conditions of the heat exchanger.

[0005] In one aspect of the present disclosure, a method for heat exchanger operating parameter optimization and dynamic response analysis is provided, the method comprising:

[0006] Acquire time series data of heat exchanger performance parameters, preprocess the time series data, and construct a dynamic response database; wherein the heat exchanger performance parameters include operating parameters and output parameters of the heat exchanger;

[0007] Based on the long short-term memory network and the deep belief network, a performance prediction model is constructed, and based on the dynamic response database, the performance prediction model is trained using a back propagation algorithm;

[0008] The trained performance prediction model is combined with a multi-objective particle swarm optimization algorithm to reversely derive the corresponding target operating parameters for the target output parameters of the heat exchanger, and perform dynamic response analysis and performance prediction of the heat exchanger under variable operating conditions to find the optimal operating parameters.

[0009] Optionally, the time series data of the operating parameters include cold and hot side inlet flow, temperature, and pressure; the time series data of the output parameters include cold and hot side heat exchange coefficient, heat exchange amount, working fluid flow field and temperature field, and rebalancing time;

[0010] The step of obtaining the time series data of the heat exchanger performance parameters includes:

[0011] Obtain the time series data in at least one of the following ways:

[0012] By actually operating the heat exchanger, recording the performance data of the heat exchanger;

[0013] Using numerical simulation to simulate the performance of the heat exchanger under different working conditions, and generate simulation data corresponding to the performance parameters of the heat exchanger;

[0014] Relevant literature and public data were consulted and analyzed, and performance data of the heat exchanger under various working conditions were collected.

[0015] Optionally, the preprocessing of the time series data includes:

[0016] Cleaning, denoising and standardizing the time series data to remove outliers and missing values ​​to obtain a corresponding data set;

[0017] According to a certain ratio, the data set is divided into a training set and a test set.

[0018] Optionally, the long short-term memory network includes three hidden layers, each of the hidden layers includes a preset number of neurons, the long short-term memory network uses an Adam optimizer, and the long short-term memory network also includes a Dropout layer;

[0019] The deep belief network includes three or four layers of restricted Boltzmann machines, and the number of neurons included in each layer of the restricted Boltzmann machine increases successively to achieve layer-by-layer greedy training.

[0020] Optionally, the long short-term memory network is used to process the time series data of the operating parameters, capture their impact on the output parameters, and analyze the time series characteristics of the output parameters;

[0021] The deep belief network is used to further perform more complex multivariate modeling and optimization based on the time series features provided by the long short-term memory network, and to model complex nonlinear relationships through a multi-layer structure.

[0022] Optionally, the training of the performance prediction model based on the dynamic response database using a back propagation algorithm includes:

[0023] The performance prediction model is trained using the training set, and during the training process, data in a validation set consisting of a portion of the data in the training set is used to monitor the training progress of the performance prediction model and prevent overfitting;

[0024] The test set is used to fine-tune the parameters of the performance prediction model using the back propagation algorithm.

[0025] Optionally, the training of the performance prediction model based on the dynamic response database using a back propagation algorithm further includes:

[0026] The relative error between the predicted value and the true value is measured using the mean absolute percentage error, where the mean absolute percentage error is expressed as:

[0027]

[0028] Wherein, MAPE represents the mean absolute percentage error; i represents the sample number; n represents the number of samples; y i Represents the true value of the i-th sample; Represents the predicted value of the i-th sample.

[0029] Optionally, the trained performance prediction model is combined with a multi-objective particle swarm optimization algorithm to reversely derive corresponding target operating parameters for the target output parameters of the heat exchanger, and perform dynamic response analysis and performance prediction of the heat exchanger under variable operating conditions to find the optimal operating parameters, including heat exchanger operating parameter optimization;

[0030] The heat exchanger operating parameter optimization includes:

[0031] Initializing the particle swarm, taking the operation parameter as an input parameter and setting its initial value;

[0032] Iteratively update the speed and position update formula of the particle swarm, continuously adjust the input parameters, and use the trained performance prediction model to predict the corresponding output parameters;

[0033] According to the set target output parameters, the difference between the output parameters of each particle and the target output parameters is evaluated, and the input parameter combination with the smallest difference is selected as the optimal target operation parameter.

[0034] Optionally, the trained performance prediction model is combined with a multi-objective particle swarm optimization algorithm to reversely derive corresponding target operating parameters for the target output parameters of the heat exchanger, and perform dynamic response analysis and performance prediction of the heat exchanger under variable operating conditions to find the optimal operating parameters, including dynamic response analysis of the heat exchanger;

[0035] The heat exchanger dynamic response analysis includes:

[0036] Inputting the initial operating parameters into the trained performance prediction model to predict the initial state;

[0037] Gradually changing the operating parameters, simulating the dynamic response of the heat exchanger under different operating conditions, recording and analyzing the changes in the output parameters of the performance prediction model under different operating conditions, and capturing the dynamic response characteristics of the heat exchanger;

[0038] By comparing the dynamic response characteristics under different operating conditions, the performance and energy efficiency of the heat exchanger under various working conditions can be evaluated.

[0039] Another aspect of the present disclosure provides a device for heat exchanger operating parameter optimization and dynamic response analysis, the device comprising:

[0040] A construction module is used to obtain time series data of heat exchanger performance parameters, preprocess the time series data, and construct a dynamic response database; wherein the heat exchanger performance parameters include operating parameters and output parameters of the heat exchanger;

[0041] A training module, used to construct a performance prediction model based on a long short-term memory network and a deep belief network, and train the performance prediction model using a back propagation algorithm based on the dynamic response database;

[0042] The prediction and analysis module is used to use the trained performance prediction model in combination with a multi-objective particle swarm optimization algorithm to reversely derive the corresponding target operating parameters for the target output parameters of the heat exchanger, and perform dynamic response analysis and performance prediction of the heat exchanger under variable operating conditions to find the optimal operating parameters.

[0043] Compared with the prior art, the present invention utilizes a long short-term memory network model and a deep belief model to perform high-precision prediction of the heat exchanger performance, and combines a multi-objective particle swarm optimization algorithm to achieve operation parameter optimization. It has significant advantages in capturing long-term and short-term dependencies and complex nonlinear relationships in time series data. By constructing a dynamic response database, it can accurately evaluate the performance of the heat exchanger under different operating conditions in a relatively short time, and find the optimal input parameter combination that meets the target output. It not only improves the accuracy and reliability of the heat exchanger performance prediction, significantly saves computing costs, but also improves the operating efficiency of the heat exchanger under various operating conditions, provides a scientific basis for performance prediction and optimization under different operating conditions, masters the operation control strategy of rapid start and stop and safe peak regulation of the heat exchanger unit, and promotes the development and application of heat exchanger technology. It has important practical significance and broad application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplifications do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0045] Figure 1 A flow chart of a method for heat exchanger operating parameter optimization and dynamic response analysis provided in one embodiment of the present disclosure;

[0046] Figure 2 A flow chart of a method for heat exchanger operating parameter optimization and dynamic response analysis provided in another embodiment of the present disclosure;

[0047] Figure 3 A schematic diagram of a numerical model of a straight-channel compact heat exchanger unit provided in accordance with another embodiment of the present disclosure;

[0048] Figure 4 A schematic structural diagram of a heat exchanger operating parameter optimization and dynamic response analysis device provided in another embodiment of the present disclosure. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical scheme and advantages of the embodiments of the present disclosure clearer, the embodiments of the present disclosure will be described in detail below in conjunction with the accompanying drawings. However, it can be understood by those skilled in the art that in each embodiment of the present disclosure, many technical details are proposed in order to enable readers to better understand the present disclosure. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed for protection in the present disclosure can also be implemented. The division of the following embodiments is for the convenience of description and should not constitute any limitation on the specific implementation of the present disclosure. The various embodiments can be combined and referenced with each other without contradiction.

[0050] One embodiment of the present disclosure relates to a method for heat exchanger operating parameter optimization and dynamic response analysis, the process of which is as follows: Figure 1 As shown, including:

[0051] Step S110, obtaining time series data of heat exchanger performance parameters, preprocessing the time series data, and constructing a dynamic response database; wherein the heat exchanger performance parameters include operating parameters and output parameters of the heat exchanger.

[0052] Specifically, step S110 is used to construct a dynamic response database. Figure 2 In order to build a database for performance prediction model training, step S110 can obtain time series data of heat exchanger performance parameters through actual operation, numerical simulation, literature research, etc. Among them, the operating parameters of the heat exchanger may include but are not limited to the working fluid inlet flow rate M, temperature T i 、Pressure P i The output parameters of the heat exchanger may include but are not limited to heat transfer Q, heat transfer coefficient h, working fluid outlet temperature T o 、Pressure P o , rebalancing time t, etc.

[0053] Step S120, constructing a performance prediction model based on the long short-term memory network and the deep belief network, and training the performance prediction model using a back propagation algorithm based on a dynamic response database.

[0054] Specifically, as advanced models in the field of deep learning, Long Short-Term Memory (LSTM) and Deep Belief Network (DBN) have significant advantages in processing time series data and complex nonlinear relationships. LSTM can effectively capture long-term and short-term dependencies in time series data, while DBN is good at modeling complex nonlinear features. Combining the two can significantly improve the prediction effect of heat exchanger performance.

[0055] Step S120 is used to build and train a performance prediction model. Figure 2 , step S120 uses the LSTM model and the DBN model to jointly construct a performance prediction model. Among them, the performance prediction model first uses the LSTM model to process the time series data of the operation parameters to generate time series features, and then inputs these time series features into the DBN model to capture and model complex nonlinear relationships through the DBN model. Step S120 also uses the back propagation algorithm to adjust the parameters of the LSTM model and the DBN model to ensure the optimal performance of the model.

[0056] Step S130, using the trained performance prediction model combined with the multi-objective particle swarm optimization algorithm, reversely derive the corresponding target operating parameters for the target output parameters of the heat exchanger, and perform dynamic response analysis and performance prediction of the heat exchanger under variable operating conditions to find the optimal operating parameters.

[0057] Specifically, step S130 is used to perform operation parameter optimization and dynamic response analysis. Figure 2 Step S130 uses the trained performance prediction model, namely the LSTM-DBN model, combined with the Multi-Objective Particle Swarm Optimization (MOPSO) algorithm to predict the target output parameters (such as heat transfer Q, working fluid outlet temperature T o 、Pressure P o etc.), and reversely deduce the corresponding operating parameters (such as working fluid inlet flow M, temperature T i 、Pressure P i etc.), and conduct dynamic response analysis and performance prediction of heat exchangers under variable operating conditions to ensure that the heat exchanger achieves optimal energy efficiency.

[0058] Exemplarily, the time series data of the operating parameters include the cold and hot side inlet flow (M c 、M h ), temperature (Tci 、T hi ), pressure (P ci , P hi ). Among them, M c is the cold side inlet flow, M h is the hot side inlet flow rate, T ci is the cold side inlet temperature, T hi is the hot side inlet temperature, P ci is the cold side inlet pressure, P hi is the hot side inlet pressure.

[0059] The time series data of the output parameters include the heat transfer coefficients of the cold and hot sides (h c 、h h ), heat transfer (Q c , Q h ), working fluid flow field and temperature field, and re-balance time t. Among them, h c is the cold side heat transfer coefficient, h h is the heat transfer coefficient on the hot side, Q c is the heat exchange on the cold side, Q h Exchange heat for the hot side.

[0060] In step S110, obtaining the time series data of the heat exchanger performance parameters is a database time series data collection step, which specifically includes:

[0061] Obtain time series data by at least one of the following methods: record the performance data of the heat exchanger by actually operating the heat exchanger; simulate the performance of the heat exchanger under different operating conditions by numerical simulation to generate simulation data corresponding to the performance parameters of the heat exchanger; consult and analyze relevant literature and public data to collect the performance data of the heat exchanger under various operating conditions.

[0062] That is to say, the sources of time series data include at least one of the following: recording the performance data of the heat exchanger by actually operating the heat exchanger; simulating the performance of the heat exchanger under different working conditions using numerical simulation software such as computational fluid dynamics (CFD) simulation software to generate corresponding simulation data; consulting and analyzing relevant literature and public data to collect the performance data of the heat exchanger under various working conditions.

[0063] Taking numerical simulation as an example, when a set of transducer performance data samples are obtained through numerical simulation, Figure 3 The straight channel compact heat exchanger unit shown in the figure is numerically simulated, a series of cold and hot side inlet flow, temperature, and pressure are set, and the second-order upwind spatial discretization method and coupled algorithm are used to solve the working fluid flow and heat transfer control equation. -6The calculation is considered to have converged. The grid independence verification is carried out. When the number of grids is higher than 1149727, the change of the export parameter value is less than 0.1%. This grid is applied to the numerical simulation process.

[0064] Exemplarily, in step S110, the time series data is preprocessed, including: cleaning, denoising and standardizing the time series data, removing outliers and missing values ​​to ensure the quality and consistency of the data, and obtaining a corresponding data set. According to a certain ratio, the data set is divided into a training set and a test set to ensure the time series characteristics and representativeness of the database, and to ensure that the training data and test data of the model do not interfere with each other.

[0065] Exemplarily, in step S120, a performance prediction model is constructed based on the long short-term memory network and the deep belief network, which belongs to the model architecture design part.

[0066] Combined Figure 2 , the LSTM network includes three hidden layers, each of which includes a preset number of neurons, for example, each hidden layer can include 20 neurons. The LSTM network uses the Adam optimizer, and the learning rate can be set to 0.001. To prevent overfitting, the LSTM network also includes a Dropout layer, and the dropout rate can be set to 0.2. The batch size can be set to 64, and the number of iterations can be set to 1000. The data set is divided into a training set and a test set in a ratio of 7:3, and 10% of the training data is used as a validation set.

[0067] Combined Figure 2 , the deep belief network includes three or four layers of restricted Boltzmann machines (RBM), and the number of neurons included in each layer of RBM increases successively to achieve layer-by-layer greedy training. Taking the deep belief network including four layers of RBM as an example, each layer of RBM can include 50, 100, 150, and 200 neurons respectively, and greedy training is performed layer by layer. The learning rate of each layer of RBM is 0.01, the batch size is 32, and the number of iterations is 500. After the pre-training is completed, the back propagation algorithm is used to fine-tune the model and further optimize the network parameters.

[0068] Exemplarily, in step S120, the performance prediction model is trained using the back propagation algorithm based on the dynamic response database, which belongs to the model training and integration part, and specifically includes: the long short-term memory network is used to process the time series data of the operating parameters, capture their impact on the output parameters, and analyze the time series characteristics of the output parameters; the deep belief network is used to further perform more complex multivariate modeling and optimization based on the time series characteristics provided by the long short-term memory network, and model complex nonlinear relationships through a multi-layer structure.

[0069] That is to say, based on the time series features provided by the long short-term memory network, step S120 further uses the deep belief network for more complex multivariable modeling and optimization, models complex nonlinear relationships through a multi-layer structure, and further improves the accuracy and reliability of the heat exchanger performance prediction.

[0070] Through training, the performance prediction model learns the relationship between operating parameters and output parameters, and is able to apply these learned relationships to make predictions when encountering new data.

[0071] Exemplarily, the model training and integration part in step S120 also includes: using the training set to train the performance prediction model, and during the training process, using the data in the validation set composed of part of the data in the training set to monitor the training progress of the performance prediction model and prevent overfitting; using the test set, using the back propagation algorithm to fine-tune the parameters of the performance prediction model. The relative error between the predicted value and the true value is measured using the mean absolute percentage error, where the expression of the mean absolute percentage error is:

[0072]

[0073] Among them, MAPE represents the mean absolute percentage error; i represents the sample number; n represents the number of samples; y i Represents the true value of the i-th sample; Represents the predicted value of the i-th sample.

[0074] Specifically, when the value of MAPE is less than 5%, it is considered that the prediction result of the performance prediction model is relatively accurate.

[0075] Exemplarily, step S130 includes heat exchanger operating parameter optimization, which uses a multi-objective particle swarm optimization (MOPSO) algorithm for parameter optimization, which uses a performance prediction model to perform multiple iterations and particle swarm updates to find the best input parameter combination that can meet the set target output parameters.

[0076] The optimization of heat exchanger operating parameters specifically includes: first, initializing the particle group, setting the operating parameters (such as the working fluid inlet flow M, temperature T i 、Pressure P i ) as the input parameter and set its initial value; secondly, it iteratively update through the speed and position update formula of the particle swarm, continuously adjust the input parameters, and use the trained performance prediction model to predict the corresponding output parameters; finally, according to the set target output parameters, evaluate the difference between the output parameters of each particle and the target output parameters, and select the input parameter combination with the smallest difference as the optimal solution, that is, the optimal target operation parameters.

[0077] Exemplarily, step S130 also includes a dynamic response analysis of the heat exchanger, specifically including: first, inputting the initial operating parameters into the trained performance prediction model to predict the initial state; then, gradually changing the operating parameters to simulate the dynamic response of the heat exchanger under different operating conditions, recording and analyzing the changes in the output parameters of the performance prediction model under different operating conditions, and capturing the dynamic response characteristics of the heat exchanger (such as temperature change curves, pressure change curves, etc.); finally, by comparing the dynamic response characteristics under different operating conditions, the performance and energy efficiency of the heat exchanger under various operating conditions are evaluated.

[0078] The heat exchanger operating parameter optimization and dynamic response analysis method provided by the embodiment of the present disclosure is an effective and innovative heat exchanger operating parameter optimization and dynamic response analysis method based on a dynamic response database compared to the prior art. The method uses LSTM model and DBN model to predict the performance of the heat exchanger with high precision, and combines the multi-objective particle swarm optimization algorithm to achieve operating parameter optimization. Compared with the traditional prediction method based on empirical formula and simplified model, the heat exchanger operating parameter optimization and dynamic response analysis method provided by the embodiment of the present disclosure has significant advantages in capturing long-term and short-term dependencies and complex nonlinear relationships in time series data. By constructing a dynamic response database, the performance of the heat exchanger under different working conditions can be accurately evaluated in a short time, and the optimal input parameter combination that meets the target output can be found. It not only improves the accuracy and reliability of the heat exchanger performance prediction, significantly saves the computing cost, but also improves the operating efficiency of the heat exchanger under various working conditions, provides a scientific basis for performance prediction and optimization under different working conditions, masters the operation control strategy of rapid start and stop and safe peak regulation of the heat exchanger unit, and promotes the development and application of heat exchanger technology, which has important practical significance and wide application value.

[0079] Another embodiment of the present disclosure relates to a heat exchanger operating parameter optimization and dynamic response analysis device, such as Figure 4 As shown, it includes a construction module 410, a training module 420, and a prediction and analysis module 430.

[0080] The construction module 410 is used to obtain the time series data of the heat exchanger performance parameters, pre-process the time series data, and construct a dynamic response database; wherein the heat exchanger performance parameters include the operating parameters and output parameters of the heat exchanger.

[0081] The training module 420 is used to construct a performance prediction model based on the long short-term memory network and the deep belief network, and train the performance prediction model based on the dynamic response database using a back propagation algorithm.

[0082] The prediction and analysis module 430 is used to utilize the trained performance prediction model in combination with the multi-objective particle swarm optimization algorithm to reversely derive the corresponding target operating parameters for the target output parameters of the heat exchanger, and to perform dynamic response analysis and performance prediction of the heat exchanger under variable operating conditions to find the optimal operating parameters.

[0083] The specific implementation method of the heat exchanger operating parameter optimization and dynamic response analysis device provided in the embodiment of the present disclosure can be found in the heat exchanger operating parameter optimization and dynamic response analysis device provided in the embodiment of the present disclosure, and will not be repeated here.

[0084] Compared with the prior art, the heat exchanger operating parameter optimization and dynamic response analysis device provided in the embodiment of the present disclosure uses LSTM model and DBN model to predict the performance of the heat exchanger with high precision, and combines the multi-objective particle swarm optimization algorithm to realize the optimization of operating parameters. It has significant advantages in capturing long-term and short-term dependencies and complex nonlinear relationships in time series data. By constructing a dynamic response database, it can accurately evaluate the performance of the heat exchanger under different working conditions in a short time, and find the best input parameter combination that meets the target output. It not only improves the accuracy and reliability of the heat exchanger performance prediction, significantly saves the computing cost, but also improves the operating efficiency of the heat exchanger under various working conditions, provides a scientific basis for performance prediction and optimization under different working conditions, masters the operation control strategy of rapid start and stop and safe peak regulation of the heat exchanger unit, and promotes the development and application of heat exchanger technology. It has important practical significance and broad application value.

[0085] Those skilled in the art will appreciate that the above-mentioned embodiments are specific embodiments for implementing the present disclosure, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present disclosure.

Claims

1. A method for heat exchanger operating parameter optimization and dynamic response analysis, characterized in that: The method comprises: Acquire time series data of heat exchanger performance parameters, preprocess the time series data, and construct a dynamic response database; wherein the heat exchanger performance parameters include operating parameters and output parameters of the heat exchanger; Based on the long short-term memory network and the deep belief network, a performance prediction model is constructed, and based on the dynamic response database, the performance prediction model is trained using a back propagation algorithm; The trained performance prediction model is combined with a multi-objective particle swarm optimization algorithm to reversely derive the corresponding target operating parameters for the target output parameters of the heat exchanger, and perform dynamic response analysis and performance prediction of the heat exchanger under variable operating conditions to find the optimal operating parameters.

2. The method according to claim 1, characterized in that The time series data of the operating parameters include the cold and hot side inlet flow, temperature, and pressure; the time series data of the output parameters include the cold and hot side heat transfer coefficient, heat transfer, working fluid flow field and temperature field, and rebalancing time; The step of obtaining the time series data of the heat exchanger performance parameters includes: Obtain the time series data in at least one of the following ways: By actually operating the heat exchanger, recording the performance data of the heat exchanger; Using numerical simulation to simulate the performance of the heat exchanger under different working conditions, and generate simulation data corresponding to the performance parameters of the heat exchanger; Relevant literature and public data were consulted and analyzed, and performance data of the heat exchanger under various working conditions were collected.

3. The method according to claim 1, characterized in that The preprocessing of the time series data includes: Cleaning, denoising and standardizing the time series data to remove outliers and missing values ​​to obtain a corresponding data set; According to a certain ratio, the data set is divided into a training set and a test set.

4. The method according to claim 1, characterized in that: The long short-term memory network includes three hidden layers, each of the hidden layers includes a preset number of neurons, the long short-term memory network uses an Adam optimizer, and the long short-term memory network also includes a Dropout layer; The deep belief network includes three or four layers of restricted Boltzmann machines, and the number of neurons included in each layer of the restricted Boltzmann machine increases successively to achieve layer-by-layer greedy training.

5. The method according to claim 4, characterized in that The LSTM network is used to process the time series data of the operating parameters, capture their influence on the output parameters, and analyze the time series characteristics of the output parameters; The deep belief network is used to further perform more complex multivariate modeling and optimization based on the time series features provided by the long short-term memory network, and to model complex nonlinear relationships through a multi-layer structure.

6. The method according to claim 3, characterized in that The method of training the performance prediction model based on the dynamic response database using a back propagation algorithm includes: The performance prediction model is trained using the training set, and during the training process, data in a validation set consisting of a portion of the data in the training set is used to monitor the training progress of the performance prediction model and prevent overfitting; The test set is used to fine-tune the parameters of the performance prediction model using the back propagation algorithm.

7. The method according to claim 6, characterized in that The training of the performance prediction model based on the dynamic response database using a back propagation algorithm also includes: The relative error between the predicted value and the true value is measured using the mean absolute percentage error, where the mean absolute percentage error is expressed as: Wherein, MAPE represents the mean absolute percentage error; i represents the sample number; n represents the number of samples; y i Represents the true value of the i-th sample; Represents the predicted value of the i-th sample.

8. The method according to claim 1, characterized in that The trained performance prediction model is combined with a multi-objective particle swarm optimization algorithm to reversely derive corresponding target operating parameters for the target output parameters of the heat exchanger, and perform dynamic response analysis and performance prediction of the heat exchanger under variable operating conditions to find the best operating parameters, including heat exchanger operating parameter optimization; The heat exchanger operating parameter optimization includes: Initialize the particle swarm, take the operation parameter as an input parameter, and set its initial value; Iteratively update the speed and position update formula of the particle swarm, continuously adjust the input parameters, and use the trained performance prediction model to predict the corresponding output parameters; According to the set target output parameters, the difference between the output parameters of each particle and the target output parameters is evaluated, and the input parameter combination with the smallest difference is selected as the optimal target operation parameter.

9. The method according to claim 1, characterized in that: The trained performance prediction model is combined with a multi-objective particle swarm optimization algorithm to reversely derive corresponding target operating parameters for the target output parameters of the heat exchanger, and perform dynamic response analysis and performance prediction of the heat exchanger under variable operating conditions to find the optimal operating parameters, including dynamic response analysis of the heat exchanger; The heat exchanger dynamic response analysis includes: Inputting the initial operating parameters into the trained performance prediction model to predict the initial state; Gradually changing the operating parameters, simulating the dynamic response of the heat exchanger under different operating conditions, recording and analyzing the changes in the output parameters of the performance prediction model under different operating conditions, and capturing the dynamic response characteristics of the heat exchanger; By comparing the dynamic response characteristics under different operating conditions, the performance and energy efficiency of the heat exchanger under various working conditions can be evaluated.

10. A device for heat exchanger operating parameter optimization and dynamic response analysis, characterized in that: The device comprises: A construction module is used to obtain time series data of heat exchanger performance parameters, preprocess the time series data, and construct a dynamic response database; wherein the heat exchanger performance parameters include operating parameters and output parameters of the heat exchanger; A training module, used to construct a performance prediction model based on a long short-term memory network and a deep belief network, and train the performance prediction model using a back propagation algorithm based on the dynamic response database; The prediction and analysis module is used to use the trained performance prediction model in combination with a multi-objective particle swarm optimization algorithm to reversely derive the corresponding target operating parameters for the target output parameters of the heat exchanger, and perform dynamic response analysis and performance prediction of the heat exchanger under variable operating conditions to find the optimal operating parameters.

Images