Heat regulation method and system for heat station of heat supply system based on secondary side flow

Abstract

The application discloses a heat regulation method for a heat supply system heat station based on secondary side flow, comprising: establishing a load prediction model for each heat station; when the heat load of each heat station changes by more than a threshold value, calculating the required secondary side water supply flow according to the heat load prediction value; establishing a plate heat exchanger model for each heat station heat exchanger; based on a double attention mechanism LSTM model, establishing a data-driven model between the adjustment parameters of the secondary side circulating pump and the auxiliary adjustment device of each heat station and the corresponding data including the primary side water supply flow and temperature, the heat exchange amount between the primary side and the secondary side based on the plate heat exchanger model; adjusting the secondary side circulating pump and the auxiliary adjustment device of each heat station according to the data-driven model to meet the required secondary side water supply flow demand value of each heat station and change the heat entering the secondary side from the primary side; correcting the adjustment parameters of the secondary side circulating pump and the auxiliary adjustment device of each heat station to obtain optimal adjustment parameters.

Description

Technical Field

[0001] This invention belongs to the field of intelligent heating regulation technology, specifically relating to a method for regulating the heat of a heating station in a heating system based on secondary side flow. Background Technology

[0002] The heating system produces high-temperature hot water in the heat source plant and drives the hot water to circulate in the primary side pipe network, delivering heat energy to various heating stations. In the heating station, heat exchange occurs between the primary and secondary sides, transferring heat from the primary side to the secondary side. The secondary side then supplies heat to various heat users in the secondary side pipe network. However, in order to ensure the heating needs of heat users and avoid insufficient heating, it is common for the secondary side circulating water flow rate to be greater than the designed circulating water flow rate in the operation of heat exchange stations in centralized heating systems. This results in excessively high secondary side return water temperature, excessive heating for heat users near the return water side, and energy waste. In addition, it will cause a decrease in the heat exchange efficiency of the heat exchanger. Usually, in order to maintain the required secondary side flow rate and supply water temperature, the primary side circulating water flow rate must be increased, which will cause hydraulic imbalance in the primary network. Therefore, how to change the heat entering the secondary side from the primary side by adjusting the secondary side flow rate without changing the primary network circulation water flow rate and temperature, and without changing the primary side hydraulic balance, and how to establish a corresponding plate heat exchange relationship between the secondary side flow rate adjustment and the primary side flow rate and temperature while ensuring the hydraulic balance of the entire network, is an urgent problem to be solved.

[0003] Based on the above technical problems, it is necessary to design a new method for heat regulation of heating stations in heating systems based on secondary side flow. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a heat regulation method for a heating station based on secondary side flow rate. This method can regulate the secondary side flow rate by means of a secondary side circulation pump and an auxiliary regulating device while keeping the primary side flow rate and temperature constant. This changes the amount of heat entering the secondary side from the primary side. In other words, the regulating action of the circulation pump and the auxiliary regulating device creates a corresponding heat exchange with the primary side flow rate and temperature, thus ensuring the heat demand of the secondary side of the heating station.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] This invention provides a method for regulating the heat output of a heating system's heat exchange station based on secondary-side flow rate. The method includes:

[0007] Step S1: Obtain historical operating data and weather data for each heating station. After performing meteorological model clustering, feature importance assessment, and model training on the obtained data through a prediction model, establish a load prediction model for each heating station.

[0008] Step S2: When the heat load change of each heating station exceeds the threshold, calculate the required secondary water supply flow rate based on the heat load forecast value;

[0009] Step S3: Establish plate heat exchanger models for each heat exchanger in the heating station;

[0010] Step S4: Based on the dual attention mechanism LSTM model, establish a data-driven model between the regulation parameters of the secondary circulating pumps and auxiliary regulation devices of each heating station and the corresponding primary side water supply flow rate and temperature, and the heat exchange between the primary and secondary sides based on the plate heat exchanger model.

[0011] Step S5: Adjust the secondary circulation pumps and auxiliary adjustment devices of each heating station according to the data-driven model to meet the secondary water supply flow requirements of each heating station and change the heat entering the secondary side from the primary side.

[0012] Step S6: Based on the established simulation model of the secondary network of the heating system, analyze the operating conditions after the secondary side flow rate is adjusted, and then correct the adjustment parameters of the secondary side circulation pumps and auxiliary adjustment devices of each heating station to obtain the optimal adjustment parameters of the secondary side circulation pumps and auxiliary adjustment devices.

[0013] Furthermore, step S1 includes:

[0014] Historical operating data and weather data are obtained from the heat meters deployed at the heating station and the weather data interface connected to the heating system, respectively. The historical operating data includes at least the secondary side water supply flow rate, secondary side supply and return water temperature, secondary side supply and return water pressure, and heat load. The weather data includes at least temperature, humidity, wind speed, and sunlight.

[0015] The historical operational data and weather data were used as model samples. Gaussian mixture model (GMM) clustering was adopted, and meteorological feature clustering analysis was performed based on the properties of the historical weather data to obtain multiple meteorological models.

[0016] The random forest algorithm is used to select features from the data in the model samples, and the feature subset with high feature importance is selected. The selected feature subset is then input into the optimized LSSVM model. After training the data under each meteorological mode, a heat station load prediction model is established. Finally, the corresponding load prediction results under each meteorological mode are superimposed to obtain the final heat station load prediction result.

[0017] Furthermore, the process of using the random forest algorithm to select features from the model samples and choosing a subset of features with higher importance includes:

[0018] Random forests consist of multiple decision trees. Feature importance is calculated based on the contribution rate of each feature in each decision tree. The feature importance of a feature is obtained by averaging its contribution rates across all decision trees. The contribution rate is calculated using the Gini coefficient. The feature importance of the j-th feature in node a is calculated based on the change in the Gini index, expressed as: VIM ja =GI a -GI b -GI c ;GI a GI b and GI c These are the Gini coefficients of node a, and the two new nodes b and c generated after the branch from node a, respectively.

[0019] Assuming there are n trees in a random forest, the importance of the j-th feature across all trees is expressed as: The sum of the importance of features across n trees; VIM ij The sum of the importance of the j-th feature in the i-th tree;

[0020] The average of the sums of the importance of the j-th feature is the feature importance of the j-th feature. Let p = 1, 2, 3, ..., n, representing the sum of the importance of m features across n trees.

[0021] By sorting all features by their importance from highest to lowest, the top n features with the highest importance are selected as a feature subset.

[0022] When performing meteorological model clustering, the GMM model employs the expectation-maximization algorithm to estimate the initial GMM's mean and covariance parameters. The expectation-maximization algorithm includes: an expectation step: setting the number of clusters in the GMM model, solving for the preliminary estimates of the initial GMM's mean and covariance, and calculating the probability that weather data belongs to only the corresponding cluster; a maximization step: using the maximum likelihood function to assign data points to clusters with higher probabilities, while simultaneously updating the GMM's mean and covariance; finally, the expectation and maximization steps are iterated until the parameters converge or the likelihood function converges, yielding the meteorological model clustering results.

[0023] The optimized LSSVM model employs a metaheuristic optimization algorithm (AOA) to optimize the kernel parameters σ and regularization parameter γ of the LSSVM model. This includes: initializing the parameters of the AOA optimization algorithm, including population size, maximum number of iterations, local development accuracy, and acceleration function; randomly generating a population, setting its initial position parameters (σ, γ), calculating individual fitness values ​​using root mean square error, and comparing the fitness values ​​to obtain the current optimal population position; determining whether the initial population has entered the exploration or development phase, and updating the initial population position; comparing the updated populations, taking the position with the lowest fitness as the optimal population position, and determining whether the iteration conditions are met; and using the optimal value generated at the end of the iteration as the LSSVM parameters (σ, γ) for model prediction.

[0024] Furthermore, step S2 includes: when the predicted heat load value of each heating station is compared with the current heat load value, if the change in heat load value exceeds the set threshold, the required secondary side water supply flow rate is calculated based on the predicted heat load value and the set value of the secondary side supply and return water temperature; otherwise, the current system operation condition is maintained.

[0025] Furthermore, in step S3, establishing a plate heat exchanger model for each heat exchanger in the heating station includes: training the plate heat exchanger model using a neural network algorithm; and establishing a heat exchanger model by fitting data of different secondary side water supply flow rates, different heat exchange rates, and different primary side water supply flow rates and temperatures. This model is used to describe the relationship between the secondary side water supply flow rate and the heat obtained from the primary side through the heat exchanger, given the primary side water supply flow rate and temperature.

[0026] Furthermore, step S4 includes:

[0027] By fitting historical data of primary side water supply flow rate and temperature, heat acquired by the secondary side from the primary side through the heat exchanger, and other relevant parameters affecting the secondary side flow rate corresponding to different secondary side circulation pumps and auxiliary regulating devices at various heating stations, a secondary side flow rate control model is established. This model is used to describe the secondary side flow rate regulation parameters required by the model output under certain model input vectors: primary side water supply flow rate and temperature, heat acquired by the secondary side from the primary side through the heat exchanger, and other relevant parameters affecting the secondary side flow rate. These parameters include the regulation parameters of the secondary side circulation pumps and auxiliary regulating devices.

[0028] The auxiliary regulating device includes a mixing device installed on the secondary side of the heating station. The flow rate of the mixing pipe is adjusted by regulating the valve opening of the mixing device, and a portion of the secondary side return water flow is introduced into the mixing pipe. At the same time, the secondary side circulation flow rate is adjusted by adjusting the parameters of the secondary side circulation pump. The secondary side supply water flow rate is regulated by the secondary side circulation pump and the mixing device together. Other parameters related to the secondary side flow rate include the secondary side supply and return water pressure difference, the secondary side circulation pump properties, the auxiliary regulating device properties, and the secondary side supply and return water pressure.

[0029] The secondary-side flow control model is trained using an LSTM model based on a dual attention mechanism, comprising: an input vector layer, a feature attention layer, an LSTM network, a temporal attention layer, and a fully connected layer output; the input vector layer obtains an input feature vector x composed of the secondary-side flow regulation history sequence and relevant input feature sequences. t The input feature vector x is then transmitted to the feature attention layer, where it is extracted through dynamic allocation of feature attention weights to obtain a weighted and corrected input feature vector x. t We construct an LSTM network layer structure and extract hidden temporal correlation information from the weighted and corrected input feature vector to obtain the hidden layer state h at each historical time step. t The temporal attention layer mines the correlation between information from each moment of the relevant feature time series and the current moment's data, and uses a temporal attention mechanism to calculate the influence weight of the output information at each historical moment. t Finally, the global hidden layer state h of the information at each historical moment is... t The input is fed into a fully connected layer, and the output is the predicted value y of the secondary side flow regulation for the next n steps. t+n .

[0030] Furthermore, the feature attention layer takes the input feature vector x as input. t As input to the feature attention mechanism, attention weights are calculated for the m features at the current time, denoted as: e t =σ(W e x t +b e );x t =[x 1,t ,x 2,t ,…,x m,t ]; e t =[e 1,t ,e 2,t ,…,e m,t ]; σ() is the Sigmoid activation function; W e b is a trainable weight matrix; e The bias vector for calculating the feature attention weights;

[0031] The feature attention weight coefficients are normalized to obtain the feature attention weights α. t =[α 1,t ,α 2,t ,…,α m,t The attention weight value for the m-th feature is expressed as:

[0032] The feature attention weight α obtained at the current time. t With input feature vector x t Calculate the weighted and corrected input feature vector x t ′, represented as: x t ′=α t ⊙x t =[α 1,t x 1,t α 2,t x 2,t … α m,t x m,t ];⊙ is for Hadama accumulation;

[0033] In addition to the feature attention mechanism, the feature attention layer also includes a CNN network, which is used to extract local features from the input data before calculating the attention weights.

[0034] Furthermore, the temporal attention layer uses the hidden layer state h of the LSTM network at time t. t As input to the temporal attention mechanism, the temporal attention weights corresponding to each historical moment at the current moment are calculated and expressed as: l t =ReLU(W d h t +b d );h t =[h 1,t ,h 2,t ,…,h k,t ]; k is the length of the input sequence time window; l t =[l 1,t ,l 2,t ,…,l k,t ]; ReLU() is the activation function; W d b is a trainable weight matrix; d This is the bias vector for calculating the temporal attention weights;

[0035] The temporal attention weight coefficients are normalized to obtain the temporal attention weight β. t =[β 1,t ,β 2,t ,…,β k,t The attention weight value at time k is expressed as:

[0036] The time attention weight β obtained at the current moment t With hidden layer state h t Calculate the global hidden layer state h t ′, represented as: It represents matrix multiplication.

[0037] Furthermore, in step S6, after analyzing the operating conditions of the secondary side flow rate after adjustment based on the established simulation model of the secondary heating system, the adjustment parameters of the secondary side circulating pumps and auxiliary adjustment devices of each heating station are corrected to obtain the optimal adjustment parameters of the secondary side circulating pumps and auxiliary adjustment devices, including:

[0038] Based on the pre-established simulation model of the secondary network of the heating system, the hydraulic conditions of the system operation after the flow rate is adjusted according to the adjustment parameters of the secondary circulation pump and the mixing device are analyzed. Based on the hydraulic misalignment, the adjustment parameters of the secondary circulation pump and the auxiliary adjustment device of each heating station are corrected to obtain the optimal adjustment parameters of the secondary circulation pump and the auxiliary adjustment device.

[0039] The hydraulic mismatch is the ratio of the actual secondary water supply flow to the predicted secondary water supply flow. If the hydraulic mismatch is greater than 1, it means that the actual secondary water supply flow of the heating station is greater than its predicted flow; if the hydraulic mismatch is less than 1, it means that the actual secondary water supply flow of the heating station is less than its predicted flow.

[0040] Based on the hydraulic imbalance, the adjustment parameters of the secondary circulating pumps and auxiliary regulating devices of each heating station are adjusted and corrected with different floating values ​​to obtain multiple regulation schemes. The regulation parameters in different regulation schemes are used as control variables. After running the regulation parameters through the constructed secondary network simulation model, the secondary side water supply flow of the heating station corresponding to different regulation schemes is output. The variance of the secondary side water supply flow of each heating station is used as the standard for hydraulic balance. The optimal regulation scheme is obtained after calculation using a global optimization algorithm.

[0041] This invention also proposes a heat regulation system for a heating system's heat station based on secondary side flow rate, the heat regulation system for the heating system's heat station comprising:

[0042] The heat load prediction module is used to acquire historical operating data and weather data of each heating station. After performing meteorological model clustering, feature importance assessment and model training on the acquired data, a load prediction model for each heating station is established.

[0043] The secondary flow demand module is used to calculate the required secondary water supply flow based on the heat load forecast when the heat load change of each heating station exceeds the threshold.

[0044] The plate heat exchanger model building module is used to build plate heat exchanger models for each heat exchanger in a heating station.

[0045] The secondary flow control module is used to establish a data-driven model based on the dual attention mechanism of the LSTM model to establish the adjustment parameters of the secondary circulating pumps and auxiliary regulating devices of each heating station and the corresponding data including the primary side water supply flow and temperature, and the heat exchange between the primary and secondary sides based on the plate heat exchanger model.

[0046] The execution module is used to adjust the secondary circulation pumps and auxiliary adjustment devices of each heating station according to the data-driven model, so as to meet the secondary water supply flow requirements of each heating station and change the heat entering the secondary side from the primary side.

[0047] The adjustment parameter correction module is used to analyze the operating conditions after the secondary side flow rate is adjusted based on the established simulation model of the secondary network of the heating system, and then correct the adjustment parameters of the secondary side circulation pumps and auxiliary adjustment devices of each heating station to obtain the optimal adjustment parameters of the secondary side circulation pumps and auxiliary adjustment devices.

[0048] The beneficial effects of this invention are:

[0049] (1) This invention acquires historical operating data and weather data from various heating stations, and establishes load prediction models for each heating station after evaluating the feature importance of the acquired data and training the model through a prediction model. When the heat load change of each heating station exceeds a threshold, the required secondary side water supply flow rate is calculated based on the predicted heat load value. Plate heat exchanger models for each heating station are established. Based on a dual attention mechanism LSTM model, a data-driven model is established between the adjustment parameters of the secondary side circulation pumps and auxiliary adjustment devices of each heating station and the corresponding primary side water supply flow rate and temperature, and the heat exchange between the primary and secondary sides based on the plate heat exchanger model. The secondary side circulation pumps and auxiliary adjustment devices of each heating station are adjusted according to the data-driven model to meet the needs of each heating station. The required secondary side water supply flow rate and the change in the heat entering the secondary side from the primary side are determined. Based on the established simulation model of the secondary heating system, the operating conditions after the secondary side flow rate adjustment are analyzed. The adjustment parameters of the secondary side circulation pumps and auxiliary adjustment devices of each heating station are then corrected to obtain the optimal adjustment parameters of the secondary side circulation pumps and auxiliary adjustment devices. Under the condition that the primary side flow rate and temperature remain unchanged and the hydraulic balance of the primary side is not altered, the secondary side flow rate can be adjusted through the secondary side circulation pumps and auxiliary adjustment devices to change the heat entering the secondary side from the primary side. In other words, the adjustment action of the circulation pumps and auxiliary adjustment devices forms a corresponding heat exchange with the primary side flow rate and temperature, ensuring the heat demand of the secondary side of the heating station.

[0050] (2) This invention evaluates the importance of features and calculates the importance scores of features in the model samples; and uses Gaussian mixture model (GMM) clustering and performs meteorological feature clustering analysis based on the properties of historical weather data to obtain multiple meteorological models; the accuracy of load forecasting can be improved through meteorological feature clustering and feature importance evaluation.

[0051] (3) This invention effectively solves the problems of LSTM models neglecting the correlation of input features in actual operation and poor performance in processing historical information by training a secondary-side flow control model using an LSTM model based on a dual attention mechanism. The feature attention mechanism mines the correlation between flow and its influencing factors, optimizes the input of the LSTM model, and improves the overall prediction accuracy; the temporal attention mechanism mines the correlation between the current flow and historical key moment information, optimizes the output of the LSTM model, and improves the prediction accuracy of key moment points.

[0052] Other features and advantages will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0053] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0054] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0055] Figure 1 This is a schematic diagram of the heat regulation method for a heating station in a heating system based on secondary side flow according to the present invention.

[0056] Figure 2 This is a flowchart of the heat station load forecasting method based on meteorological models and feature importance of the present invention;

[0057] Figure 3 This is a schematic diagram of the LSTM model structure based on the dual attention mechanism of the present invention;

[0058] Figure 4 This is a schematic diagram of the heat regulation system structure of a heating station based on secondary side flow rate according to the present invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] Example 1

[0061] Figure 1 This is a schematic diagram of a heat regulation method for a heating station in a heating system based on secondary side flow, which is involved in this invention.

[0062] like Figure 1 As shown, this embodiment 1 provides a method for regulating the heat output of a heating system's heat exchange station based on secondary side flow rate. The method includes:

[0063] Step S1: Obtain historical operating data and weather data for each heating station. After performing meteorological model clustering, feature importance assessment, and model training on the obtained data through a prediction model, establish a load prediction model for each heating station.

[0064] Step S2: When the heat load change of each heating station exceeds the threshold, calculate the required secondary water supply flow rate based on the heat load forecast value;

[0065] Step S3: Establish plate heat exchanger models for each heat exchanger in the heating station;

[0066] Step S4: Based on the dual attention mechanism LSTM model, establish a data-driven model between the regulation parameters of the secondary circulating pumps and auxiliary regulation devices of each heating station and the corresponding primary side water supply flow rate and temperature, and the heat exchange between the primary and secondary sides based on the plate heat exchanger model.

[0067] Step S5: Adjust the secondary circulation pumps and auxiliary adjustment devices of each heating station according to the data-driven model to meet the secondary water supply flow requirements of each heating station and change the heat entering the secondary side from the primary side.

[0068] Step S6: Based on the established simulation model of the secondary network of the heating system, analyze the operating conditions after the secondary side flow rate is adjusted, and then correct the adjustment parameters of the secondary side circulation pumps and auxiliary adjustment devices of each heating station to obtain the optimal adjustment parameters of the secondary side circulation pumps and auxiliary adjustment devices.

[0069] It should be noted that the secondary side water supply flow rate is a flow range with upper and lower limits for flow regulation. Usually, the minimum flow rate under hydraulic balance can be selected. By regulating the secondary side flow rate, the amount of heat entering the secondary side from the primary side can be changed without changing the hydraulic balance state of the primary side or the primary side water supply flow rate and temperature.

[0070] Figure 2 This is a flowchart of the heat station load forecasting method based on meteorological models and feature importance involved in this invention.

[0071] like Figure 2 As shown, in this embodiment, step S1 includes:

[0072] Historical operating data and weather data are obtained from the heat meters deployed at the heating station and the weather data interface connected to the heating system, respectively. The historical operating data includes at least the secondary side water supply flow rate, secondary side supply and return water temperature, secondary side supply and return water pressure, and heat load. The weather data includes at least temperature, humidity, wind speed, and sunlight.

[0073] The historical operational data and weather data were used as model samples. Gaussian mixture model (GMM) clustering was adopted, and meteorological feature clustering analysis was performed based on the properties of the historical weather data to obtain multiple meteorological models.

[0074] The random forest algorithm is used to select features from the data in the model samples, and the feature subset with high feature importance is selected. The selected feature subset is then input into the optimized LSSVM model. After training the data under each meteorological mode, a heat station load prediction model is established. Finally, the corresponding load prediction results under each meteorological mode are superimposed to obtain the final heat station load prediction result.

[0075] In this embodiment, the process of using the random forest algorithm to select features from the model samples and choosing a subset of features with higher importance includes:

[0076] Random forests consist of multiple decision trees. Feature importance is calculated based on the contribution rate of each feature in each decision tree. The feature importance of a feature is obtained by averaging its contribution rates across all decision trees. The contribution rate is calculated using the Gini coefficient. The feature importance of the j-th feature in node a is calculated based on the change in the Gini index, expressed as: VIM ja =GI a -GI b -GI c ;GI a GI b and GI c These are the Gini coefficients of node a, and the two new nodes b and c generated after the branch from node a, respectively.

[0077] Assuming there are n trees in a random forest, the importance of the j-th feature across all trees is expressed as: The sum of the importance of features across n trees; VIM ij The sum of the importance of the j-th feature in the i-th tree;

[0078] The average of the sums of the importance of the j-th feature is the feature importance of the j-th feature. Let p = 1, 2, 3, ..., n, representing the sum of the importance of m features across n trees.

[0079] By sorting all features by their importance from highest to lowest, the top n features with the highest importance are selected as a feature subset.

[0080] When performing meteorological model clustering, the GMM model employs the expectation-maximization algorithm to estimate the initial GMM's mean and covariance parameters. The expectation-maximization algorithm includes: an expectation step: setting the number of clusters in the GMM model, solving for the preliminary estimates of the initial GMM's mean and covariance, and calculating the probability that weather data belongs to only the corresponding cluster; a maximization step: using the maximum likelihood function to assign data points to clusters with higher probabilities, while simultaneously updating the GMM's mean and covariance; finally, the expectation and maximization steps are iterated until the parameters converge or the likelihood function converges, yielding the meteorological model clustering results.

[0081] After obtaining multiple meteorological models, the XGBoost learning algorithm is used to evaluate the feature importance of weather data and operational data under each meteorological model and calculate the feature importance score. Then, the selected feature data is input into the optimized LSSVM model. After training the data under each meteorological model, a heat station load prediction model is established. Finally, the corresponding load prediction results under each meteorological model are superimposed to obtain the final heat station load prediction result.

[0082] The optimized LSSVM model employs a metaheuristic optimization algorithm (AOA) to optimize the kernel parameters σ and regularization parameter γ of the LSSVM model. This includes: initializing the parameters of the AOA optimization algorithm, including population size, maximum number of iterations, local development accuracy, and acceleration function; randomly generating a population, setting its initial position parameters (σ, γ), calculating individual fitness values ​​using root mean square error, and comparing the fitness values ​​to obtain the current optimal population position; determining whether the initial population has entered the exploration or development phase, and updating the initial population position; comparing the updated populations, taking the position with the lowest fitness as the optimal population position, and determining whether the iteration conditions are met; and using the optimal value generated at the end of the iteration as the LSSVM parameters (σ, γ) for model prediction.

[0083] It should be noted that Gaussian Mixture Model (GMM) clustering uses a multidimensional Gaussian Mixture Model (GMM) to characterize the clusters of each sample. The resulting clusters are a series of probability values, with each individual in the sample having a probability corresponding to a different category. The category with the highest probability is selected as the classification criterion. Weather data includes at least temperature, humidity, wind speed, and sunshine duration. Based on the inherent properties of historical weather elements, the GMM classifies them into multiple meteorological models, enabling refined grouping training and forecasting. From a forecasting perspective, different meteorological models consider the differences in heat load forecasts under various meteorological conditions, and each model belongs to a multivariate Gaussian distribution with different parameters. For example, meteorological model 1 represents moderate wind speed, moderate to high temperature, moderate to high sunshine duration, and moderate to high humidity; meteorological model 2 represents moderate to low wind speed, moderate to low temperature, moderate to low humidity, and moderate to low sunshine duration; and meteorological model 3 represents moderate wind speed, low temperature, moderate to low humidity, and low sunshine duration.

[0084] AOA optimization algorithms include:

[0085] (1) The optimization strategy is selected by accelerating the function MOA through the mathematical optimizer, including the global exploration stage and the local development stage. When the random number r1 > MOA, the global exploration stage is carried out, and when the random number r1 < MOA, the local development stage is carried out. M max M min These represent the maximum and minimum values ​​of MOA, respectively; t is the current iteration number; T is the total number of iterations;

[0086] (2) During the exploration phase, global exploration is achieved through multiplication and division strategies to enhance global optimization capabilities and overcome premature convergence. When the random number r2 < 0.5, the division exploration strategy is executed, and when the random number r2 ≥ 0.5, the multiplication exploration strategy is executed.

[0087]

[0088]

[0089] ξ is the minimum value; u is the control parameter for adjusting the search process; UB and LB are the upper and lower boundaries of the variables; MOP is the mathematical optimizer probability;

[0090] (3) During the development phase, local exploration is achieved through addition and subtraction strategies to enhance the local optimization capability. When the random number r3 < 0.5, the subtraction exploration strategy is executed, and when the random number r3 ≥ 0.5, the addition exploration strategy is executed.

[0091]

[0092] By using AOA, which has strong optimization capabilities and fast convergence speed, to optimize two parameters of the LSSVM model, the problem of the difficulty in determining LSSVM parameters is solved. At the same time, the optimization performance is better than previous optimization algorithms, which can significantly improve the accuracy of prediction.

[0093] In this embodiment, step S2 includes: when the predicted heat load value of each heating station is compared with the current heat load value, if the change in heat load value exceeds the set threshold, the required secondary side water supply flow rate is calculated based on the predicted heat load value and the set value of the secondary side supply and return water temperature; otherwise, the current system operation condition is maintained.

[0094] In this embodiment, step S3 involves establishing a plate heat exchanger model for each heat exchanger in the heating station. This includes training the plate heat exchanger model using a neural network algorithm and establishing a heat exchanger model by fitting data on different secondary side water flow rates, different heat exchange rates, and different primary side water flow rates and temperatures. This model describes the relationship between the secondary side water flow rate and the heat obtained from the primary side through the heat exchanger, given the primary side water flow rate and temperature.

[0095] Figure 3 This is a schematic diagram of the LSTM model structure based on the dual attention mechanism involved in this invention.

[0096] like Figure 3 As shown, in this embodiment, step S4 includes:

[0097] By fitting historical data of primary side water supply flow rate and temperature, heat acquired by the secondary side from the primary side through the heat exchanger, and other relevant parameters affecting the secondary side flow rate corresponding to different secondary side circulation pumps and auxiliary regulating devices at various heating stations, a secondary side flow rate control model is established. This model is used to describe the secondary side flow rate regulation parameters required by the model output under certain model input vectors: primary side water supply flow rate and temperature, heat acquired by the secondary side from the primary side through the heat exchanger, and other relevant parameters affecting the secondary side flow rate. These parameters include the regulation parameters of the secondary side circulation pumps and auxiliary regulating devices.

[0098] The auxiliary regulating device includes a mixing device installed on the secondary side of the heating station. The flow rate of the mixing pipe is adjusted by regulating the valve opening of the mixing device, and a portion of the secondary side return water flow is introduced into the mixing pipe. At the same time, the secondary side circulation flow rate is adjusted by adjusting the parameters of the secondary side circulation pump. The secondary side supply water flow rate is regulated by the secondary side circulation pump and the mixing device together. Other parameters related to the secondary side flow rate include the secondary side supply and return water pressure difference, the secondary side circulation pump properties, the auxiliary regulating device properties, and the secondary side supply and return water pressure.

[0099] The secondary-side flow control model is trained using an LSTM model based on a dual attention mechanism, comprising: an input vector layer, a feature attention layer, an LSTM network, a temporal attention layer, and a fully connected layer output; the input vector layer obtains an input feature vector x composed of the secondary-side flow regulation history sequence and relevant input feature sequences. t The input feature vector x is then transmitted to the feature attention layer, where it is extracted through dynamic allocation of feature attention weights to obtain a weighted and corrected input feature vector x. t We construct an LSTM network layer structure and extract hidden temporal correlation information from the weighted and corrected input feature vector to obtain the hidden layer state h at each historical time step. t The temporal attention layer mines the correlation between information from each moment of the relevant feature time series and the current moment's data, and uses a temporal attention mechanism to calculate the influence weight of the output information at each historical moment. t Finally, the global hidden layer state h of the information at each historical moment is... t The input is fed into a fully connected layer, and the output is the predicted value y of the secondary side flow regulation for the next n steps. t+n .

[0100] In this embodiment, the feature attention layer takes the input feature vector x as input. t As input to the feature attention mechanism, attention weights are calculated for the m features at the current time, denoted as: e t =σ(W e x t +b e );x t =[x 1,t ,x 2,t ,…,x m,t ]; e t =[e 1,t ,e 2,t ,…,e m,t ]; σ() is the Sigmoid activation function; W e b is a trainable weight matrix; e The bias vector for calculating the feature attention weights;

[0101] The feature attention weight coefficients are normalized to obtain the feature attention weights α. t =[α 1,t ,α 2,t ,…,α m,t The attention weight value for the m-th feature is expressed as:

[0102] The feature attention weight α obtained at the current time. t With input feature vector x tCalculate the weighted and corrected input feature vector x t ′, represented as: x t ′=α t ⊙x t =[α 1,t x 1,t α 2,t x 2,t … α m,t x m,t ];⊙ is for Hadama accumulation;

[0103] In addition to the feature attention mechanism, the feature attention layer also includes a CNN network, which is used to extract local features from the input data before calculating the attention weights.

[0104] In this embodiment, the temporal attention layer uses the hidden layer state h of the LSTM network at time t. t As input to the temporal attention mechanism, the temporal attention weights corresponding to each historical moment at the current moment are calculated and expressed as: l t =ReLU(W d h t +b d );h t =[h 1,t ,h 2,t ,…,h k,t ]; k is the length of the input sequence time window; l t =[l 1,t ,l 2,t ,…,l k,t ]; ReLU() is the activation function; W d b is a trainable weight matrix; d This is the bias vector for calculating the temporal attention weights;

[0105] The temporal attention weight coefficients are normalized to obtain the temporal attention weight β. t =[β 1,t ,β 2,t ,…,β k,t The attention weight value at time k is expressed as:

[0106] The time attention weight β obtained at the current moment t With hidden layer state h t Calculate the global hidden layer state h t ′, represented as: It represents matrix multiplication.

[0107] It's important to note that the attention mechanism is a model that simulates human brain attention. It draws on the characteristic that the human brain concentrates its attention on specific areas at a particular moment, reducing or even ignoring attention to other parts. Attention assigns different weights to the model's input features, highlighting more critical influencing factors and thus helping the model make more accurate judgments. In secondary-side traffic prediction, the feature attention mechanism analyzes the importance of different input parameters to the prediction information, quantifies the weight of the influence of input features on secondary-side traffic, emphasizes key features, and weakens features with lower relevance. Finally, in the secondary-side traffic prediction, the temporal attention mechanism fully considers the significant influence of historical states on secondary-side traffic and the varying influence of secondary-side traffic at different times. By quantifying the impact of state information at each historical moment on the current traffic prediction result, it adaptively processes historical state information and strengthens the influence of state information at relevant moments.

[0108] In this embodiment, in step S6, after analyzing the operating conditions after secondary-side flow adjustment based on the established simulation model of the secondary heating system, the adjustment parameters of the secondary-side circulating pumps and auxiliary adjustment devices of each heating station are corrected to obtain the optimal adjustment parameters of the secondary-side circulating pumps and auxiliary adjustment devices, including:

[0109] Based on the pre-established simulation model of the secondary network of the heating system, the hydraulic conditions of the system operation after the flow rate is adjusted according to the adjustment parameters of the secondary circulation pump and the mixing device are analyzed. Based on the hydraulic misalignment, the adjustment parameters of the secondary circulation pump and the auxiliary adjustment device of each heating station are corrected to obtain the optimal adjustment parameters of the secondary circulation pump and the auxiliary adjustment device.

[0110] The hydraulic mismatch is the ratio of the actual secondary water supply flow to the predicted secondary water supply flow. If the hydraulic mismatch is greater than 1, it means that the actual secondary water supply flow of the heating station is greater than its predicted flow; if the hydraulic mismatch is less than 1, it means that the actual secondary water supply flow of the heating station is less than its predicted flow.

[0111] Based on the hydraulic imbalance, the adjustment parameters of the secondary circulating pumps and auxiliary regulating devices of each heating station are adjusted and corrected with different floating values ​​to obtain multiple regulation schemes. The regulation parameters in different regulation schemes are used as control variables. After running the regulation parameters through the constructed secondary network simulation model, the secondary side water supply flow of the heating station corresponding to different regulation schemes is output. The variance of the secondary side water supply flow of each heating station is used as the standard for hydraulic balance. The optimal regulation scheme is obtained after calculation using a global optimization algorithm.

[0112] Example 2

[0113] Figure 4This is a schematic diagram of the heat regulation system structure of a heating station based on secondary side flow, which is involved in this invention.

[0114] like Figure 4 As shown in Embodiment 2, a heat regulation system for a heating station based on secondary side flow is proposed. The heat regulation system for the heating station includes:

[0115] The heat load prediction module is used to acquire historical operating data and weather data of each heating station. After performing meteorological model clustering, feature importance assessment and model training on the acquired data, a load prediction model for each heating station is established.

[0116] The secondary flow demand module is used to calculate the required secondary water supply flow based on the heat load forecast when the heat load change of each heating station exceeds the threshold.

[0117] The plate heat exchanger model building module is used to build plate heat exchanger models for each heat exchanger in a heating station.

[0118] The secondary flow control module is used to establish a data-driven model based on the dual attention mechanism of the LSTM model to establish the adjustment parameters of the secondary circulating pumps and auxiliary regulating devices of each heating station and the corresponding data including the primary side water supply flow and temperature, and the heat exchange between the primary and secondary sides based on the plate heat exchanger model.

[0119] The execution module is used to adjust the secondary circulation pumps and auxiliary adjustment devices of each heating station according to the data-driven model, so as to meet the secondary water supply flow requirements of each heating station and change the heat entering the secondary side from the primary side.

[0120] The adjustment parameter correction module is used to analyze the operating conditions after the secondary side flow rate is adjusted based on the established simulation model of the secondary network of the heating system, and then correct the adjustment parameters of the secondary side circulation pumps and auxiliary adjustment devices of each heating station to obtain the optimal adjustment parameters of the secondary side circulation pumps and auxiliary adjustment devices.

[0121] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can also be implemented in other ways. The system embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0122] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0123] If the functionality is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0124] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for regulating the heat output of a heating system's heat station based on secondary-side flow, characterized in that, The heat regulation method for the heating system's heat station includes: Step S1: Obtain historical operating data and weather data for each heating station. After performing meteorological model clustering, feature importance assessment, and model training on the obtained data through a prediction model, establish a load prediction model for each heating station. Step S1 includes: Historical operating data and weather data are obtained from the heat meters deployed at the heating station and the weather data interface connected to the heating system, respectively. The historical operating data includes at least the secondary side water supply flow rate, secondary side supply and return water temperature, secondary side supply and return water pressure, and heat load. The weather data includes at least temperature, humidity, wind speed, and sunlight. The historical operational data and weather data were used as model samples. Gaussian mixture model (GMM) clustering was adopted, and meteorological feature clustering analysis was performed based on the properties of the historical weather data to obtain multiple meteorological models. The random forest algorithm is used to select features from the data in the model samples, and the feature subset with high feature importance is selected. The selected feature subset is then input into the optimized LSSVM model. After training the data under each meteorological mode, a heat station load prediction model is established. Finally, the corresponding load prediction results under each meteorological mode are superimposed to obtain the final heat station load prediction result. Step S2: When the heat load change of each heating station exceeds the threshold, calculate the required secondary water supply flow rate based on the heat load forecast value; Step S3: Establish plate heat exchanger models for each heat exchanger in the heating station; Step S4: Based on the dual attention mechanism LSTM model, establish a data-driven model between the regulation parameters of the secondary circulating pumps and auxiliary regulation devices of each heating station and the corresponding primary side water supply flow rate and temperature, and the heat exchange between the primary and secondary sides based on the plate heat exchanger model. Step S5: Adjust the secondary circulation pumps and auxiliary adjustment devices of each heating station according to the data-driven model to meet the secondary water supply flow requirements of each heating station and change the heat entering the secondary side from the primary side. Step S6: Based on the established simulation model of the secondary network of the heating system, analyze the operating conditions after the secondary side flow rate is adjusted, and then correct the adjustment parameters of the secondary side circulation pumps and auxiliary adjustment devices of each heating station to obtain the optimal adjustment parameters of the secondary side circulation pumps and auxiliary adjustment devices.

2. The heat regulation method for a heating system heat station according to claim 1, characterized in that, The process of using the random forest algorithm to select features from the model samples and choosing a subset of features with higher importance includes: Random forests consist of multiple decision trees. Feature importance is calculated based on the contribution rate of each feature in each decision tree. The feature importance of a feature is obtained by averaging its contribution rates across all decision trees. The contribution rate is calculated using the Gini coefficient. The feature importance of the j-th feature in node a is calculated based on the change in the Gini index, expressed as: ; , and These are the Gini coefficients of node a, and the two new nodes b and c generated after the branch from node a, respectively. Assuming there are n trees in a random forest, the importance of the j-th feature across all trees is expressed as: ; This represents the sum of the importance of features across the n trees. The sum of the importance of the j-th feature in the i-th tree; The average of the sums of the importance of the j-th feature is the feature importance of the j-th feature. ; Let m be the sum of the importance of all features across n trees. ; By sorting all features by their importance from highest to lowest, the top n features with the highest importance are selected as a feature subset. When performing meteorological model clustering, the GMM model employs the expectation-maximization algorithm to estimate the initial GMM's mean and covariance parameters. The expectation-maximization algorithm includes: an expectation step: setting the number of clusters in the GMM model, solving for the preliminary estimates of the initial GMM's mean and covariance, and calculating the probability that weather data belongs to only the corresponding cluster; a maximization step: using the maximum likelihood function to assign data points to clusters with higher probabilities, while simultaneously updating the GMM's mean and covariance; finally, the expectation and maximization steps are iterated until the parameters converge or the likelihood function converges, yielding the meteorological model clustering results. The optimized LSSVM model is obtained by using the metaheuristic optimization algorithm AOA to optimize the kernel parameters of the LSSVM model. and regularization parameters Optimization includes: initializing the parameters of the AOA optimization algorithm, including population size, maximum number of iterations, local exploitation accuracy, and speedup function; randomly generating the population and setting its initial position parameters. The algorithm calculates the fitness value of each individual using the root mean square error, then compares the fitness values ​​to obtain the current optimal population position; it determines whether the initial population has entered the exploration or development phase and updates the initial population position accordingly; it compares the updated populations, taking the position with the lowest fitness as the optimal population position, and determines whether the iteration conditions are met; the optimal value generated at the end of the iteration is used as the parameters of LSSVM. Perform model predictions.

3. The heat regulation method for a heating system heat station according to claim 1, characterized in that, Step S2 includes: when the predicted heat load value of each heating station is compared with the current heat load value, if the change in heat load value exceeds the set threshold, the required secondary side water supply flow rate is calculated based on the predicted heat load value and the set value of the secondary side supply and return water temperature; otherwise, the current system operation condition is maintained.

4. The method for regulating the heat output of a heating system's heat exchange station according to claim 1, characterized in that, In step S3, the plate heat exchanger model of each heat exchanger in the heating station is established, including: training the plate heat exchanger model using a neural network algorithm, and establishing the heat exchanger ...

5. The method for regulating the heat output of a heating system's heat exchange station according to claim 1, characterized in that, Step S4 includes: By fitting historical data of primary side water supply flow rate and temperature, heat acquired by the secondary side from the primary side through the heat exchanger, and other relevant parameters affecting the secondary side flow rate corresponding to different secondary side circulation pumps and auxiliary regulating devices at various heating stations, a secondary side flow rate control model is established. This model is used to describe the secondary side flow rate regulation parameters required by the model output under certain model input vectors: primary side water supply flow rate and temperature, heat acquired by the secondary side from the primary side through the heat exchanger, and other relevant parameters affecting the secondary side flow rate. These parameters include the regulation parameters of the secondary side circulation pumps and auxiliary regulating devices. The auxiliary regulating device includes a mixing device installed on the secondary side of the heating station. The flow rate of the mixing pipe is adjusted by regulating the valve opening of the mixing device, and a portion of the secondary side return water flow is introduced into the mixing pipe. At the same time, the secondary side circulation flow rate is adjusted by adjusting the parameters of the secondary side circulation pump. The secondary side supply water flow rate is regulated by the secondary side circulation pump and the mixing device together. Other parameters related to the secondary side flow rate include the secondary side supply and return water pressure difference, the secondary side circulation pump properties, the auxiliary regulating device properties, and the secondary side supply and return water pressure. The secondary-side flow control model is trained using an LSTM model based on a dual attention mechanism, comprising: an input vector layer, a feature attention layer, an LSTM network, a temporal attention layer, and a fully connected layer output; the input vector layer acquires an input feature vector composed of the secondary-side flow regulation historical sequence and relevant input feature sequences. The input feature vector is then transmitted to the feature attention layer, where it is extracted through dynamic allocation of feature attention weights to obtain a weighted and corrected input feature vector. Construct an LSTM network layer structure and extract hidden temporal correlation information from the weighted and corrected input feature vector to obtain the hidden layer state at each historical time step. The temporal attention layer mines the correlation between relevant feature time series information at each moment and the current moment data, and uses a temporal attention mechanism to calculate the influence weight of the output information at each historical moment. Finally, the global hidden layer state of information at each historical moment is... The input is fed into a fully connected layer, and the output is a predicted value of the secondary side flow regulation for the next n steps. .

6. The heat regulation method for a heating system heat station according to claim 5, characterized in that, The feature attention layer takes the input feature vector as input. As input to the feature attention mechanism, attention weights are calculated for the m features at the current time, as follows: ; ; ; Use the Sigmoid activation function; The weight matrix is ​​trainable. The bias vector for calculating the feature attention weights; The attention weight coefficients of each feature are normalized to obtain the feature attention weights. The attention weight value for the m-th feature is expressed as: ; Feature attention weights obtained at the current moment With input feature vector Calculate the weighted and corrected input feature vector , represented as: ;⊙ is for Hadama accumulation; In addition to the feature attention mechanism, the feature attention layer also includes a CNN network, which is used to extract local features from the input data before calculating the attention weights.

7. The heat regulation method for a heating system heat station according to claim 5, characterized in that, The temporal attention layer uses the hidden layer state of the LSTM network at time t. As input to the temporal attention mechanism, the temporal attention weights corresponding to each historical moment at the current moment are calculated and expressed as follows: ; ; The length of the input sequence time window; ; For activation functions; The weight matrix is ​​trainable. This is the bias vector for calculating the temporal attention weights; The temporal attention weight coefficients are normalized to obtain the temporal attention weights. The attention weight value at time k is expressed as: ; Time attention weight obtained at the current moment With hidden layer state Calculate the global hidden layer state , represented as: ; It represents matrix multiplication.

8. The heat regulation method for a heating system heat station according to claim 1, characterized in that, In step S6, after analyzing the operating conditions after secondary-side flow adjustment based on the established simulation model of the secondary heating system, the adjustment parameters of the secondary-side circulating pumps and auxiliary adjustment devices of each heating station are corrected to obtain the optimal adjustment parameters of the secondary-side circulating pumps and auxiliary adjustment devices, including: Based on the pre-established simulation model of the secondary network of the heating system, the hydraulic conditions of the system operation after the flow rate is adjusted according to the adjustment parameters of the secondary circulation pump and the mixing device are analyzed. Based on the hydraulic misalignment, the adjustment parameters of the secondary circulation pump and the auxiliary adjustment device of each heating station are corrected to obtain the optimal adjustment parameters of the secondary circulation pump and the auxiliary adjustment device. The hydraulic mismatch is the ratio of the actual secondary water supply flow to the predicted secondary water supply flow. If the hydraulic mismatch is greater than 1, it means that the actual secondary water supply flow of the heating station is greater than its predicted flow; if the hydraulic mismatch is less than 1, it means that the actual secondary water supply flow of the heating station is less than its predicted flow. Based on the hydraulic imbalance, the adjustment parameters of the secondary circulating pumps and auxiliary regulating devices of each heating station are adjusted and corrected with different floating values ​​to obtain multiple regulation schemes. The regulation parameters in different regulation schemes are used as control variables. After running the regulation parameters through the constructed secondary network simulation model, the secondary side water supply flow of the heating station corresponding to different regulation schemes is output. The variance of the secondary side water supply flow of each heating station is used as the standard for hydraulic balance. The optimal regulation scheme is obtained after calculation using a global optimization algorithm.

Images