Heat supply energy-saving analysis method and system based on data analysis
By acquiring multi-dimensional data of the heating area, deep learning models and deep reinforcement learning algorithms are used to generate heating energy-saving control actions, which solves the problem of insufficient accuracy in traditional heating control methods and achieves a balance between efficient heating energy saving and user comfort.
Patent Information
- Application Number
- CN202511093867.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-06
- Publication Date
- 2025-11-21
AI Technical Summary
Traditional heating control methods are difficult to accurately match actual energy demand, resulting in energy waste and poor user comfort. Existing methods based on statistical models or shallow machine learning have insufficient prediction accuracy and optimization capabilities when dealing with heating systems.
A data analysis-based approach is adopted to acquire multi-dimensional data of the heating area, use a deep learning model to predict energy consumption, and combine deep reinforcement learning algorithms to generate heating energy-saving control actions, which are then transmitted to the heating system controller for precise control.
It significantly improves the accuracy of energy consumption forecasting, generates better heating strategies, avoids energy waste, maintains long-term energy-saving effects, and ensures user comfort.
Smart Images
Figure CN120995008A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of heating energy conservation technology, specifically relating to a heating energy conservation analysis method and system based on data analysis. Background Technology
[0002] With the increasing energy consumption and ever-increasing environmental protection requirements, energy-saving optimization of building heating systems has become a crucial aspect of energy conservation and emission reduction. Traditional heating energy-saving methods often rely on experience-based settings, simple temperature adjustments, or rough predictions based on historical average data, which frequently fail to accurately match actual energy demand, leading to energy waste and poor user comfort. For example, fixed-time heating and simple feedback control cannot cope with the influence of multiple complex factors such as weather changes, differences in building characteristics, and user behavior habits. In recent years, although some methods based on statistical models or shallow machine learning have been introduced, their prediction accuracy and optimization capabilities still need improvement when dealing with heating system data involving a large number of nonlinear, time-varying, and strongly coupled factors. Summary of the Invention
[0003] This invention provides a data analysis-based heating energy-saving analysis method and system to solve the problem that the existing technology has low heating control accuracy, resulting in large energy waste.
[0004] On the one hand, the present invention provides a data analysis-based method for energy-saving heating analysis, comprising: Acquire multi-dimensional data on target energy consumption related to the heating area, and use a pre-deployed energy consumption analysis model to identify the multi-dimensional data on target energy consumption and determine the energy consumption prediction results; Based on the energy consumption prediction results, a control action state space is constructed, and a pre-deployed control action prediction model is used to identify the control action data and determine the heating energy-saving control action. The heating energy-saving control action is transmitted to the controller of the heating system so that the controller of the heating system can perform heating energy-saving control.
[0005] In one possible implementation, multi-dimensional data on target energy consumption related to the heating area are acquired, including: By analyzing water supply temperature, return water temperature, outdoor temperature, humidity, wind speed, solar radiation intensity, building load characteristic parameters, historical user heating behavior data, historical heating supply, electricity consumption, and water consumption, multi-dimensional data on target energy consumption related to the heating area can be obtained.
[0006] In one possible implementation, a pre-deployed energy consumption analysis model is used to identify the multi-dimensional data of the target energy consumption and determine the energy consumption prediction results, including: The target energy consumption multi-dimensional data is normalized to obtain normalized target energy consumption multi-dimensional data. The normalized multi-dimensional target energy consumption data is input into the pre-deployed energy consumption analysis model to obtain energy consumption prediction results.
[0007] In one possible implementation, the pre-deployment method of the energy consumption analysis model includes: A convolutional neural network is used to construct an energy consumption analysis model, and the model parameters of the energy consumption analysis model are initialized to obtain multiple different individual parameters; Acquire historical energy consumption multi-dimensional data and the corresponding historical energy consumption data, and obtain the fitness of each individual parameter based on the historical energy consumption multi-dimensional data and the corresponding historical energy consumption data. Based on the fitness of the individual parameter, an adaptive neighborhood search is performed on the individual parameter to obtain the individual parameter after the adaptive neighborhood search. Perform a joint search on the individual parameters after the adaptive neighborhood search to obtain the individual parameters after the joint search. Perform positional fluctuation search on the parameter individuals after the joint search of the individuals to obtain the parameter individuals after the positional fluctuation search; The parameter individuals after the position fluctuation search are subjected to information fusion mutation search to obtain the parameter individuals after information fusion mutation search. Determine if the current number of training iterations is greater than or equal to the maximum number of training iterations. If so, determine the target optimal individual based on the parameter individuals after information fusion mutation search. Otherwise, return to the step of obtaining the fitness of each parameter individual. The model parameters in the target optimal individual are used as the final model parameters of the energy consumption analysis model, and the energy consumption analysis model is pre-deployed.
[0008] In one possible implementation, a control action state space is constructed based on the energy consumption prediction results, including: The energy consumption prediction results, current water supply temperature, current return water temperature, outdoor temperature, upper limit of pipeline pressure, lower limit of pipeline pressure, upper limit of heat source output, and lower limit of heat source output are used to construct the control action state space.
[0009] In one possible implementation, a pre-deployed control action prediction model is used to identify the control action data and determine the heating energy-saving control actions, including: The control action state space is input into a pre-deployed control action prediction model to obtain the heating energy-saving control action output by the pre-deployed control action prediction model; wherein, the heating energy-saving control action includes at least the actual output of the heat source, the supply and return water temperature of the heat exchange station secondary network, the actual output, the supply and return water temperature of the heat exchange station secondary network, the valve opening of the heating pipeline, and the actual speed of the hot water pump.
[0010] In one possible implementation, the pre-deployed control action prediction model is set up as a policy network.
[0011] In one possible implementation, after determining the heating energy-saving control action, the method further includes: Obtain the reward corresponding to the heating energy-saving control action, and construct historical training experience based on the control action state space, the heating energy-saving control action, the reward, and the control action state space after executing the heating energy-saving control action; Based on the historical training experience, the strategy network is updated in conjunction with the value network to obtain the updated strategy network. The updated strategy network is then used to determine the heating energy-saving control actions in the subsequent process.
[0012] In one possible implementation, the pre-deployment method of the control action prediction model includes: pre-deploying the control action prediction model using the DQN algorithm.
[0013] On the other hand, the present invention provides a heating energy-saving analysis system based on data analysis, including: an energy consumption prediction module, a control action acquisition module, and an energy-saving control module; The energy consumption prediction module is used to acquire multi-dimensional data of target energy consumption related to the heating area, and to identify the multi-dimensional data of target energy consumption using a pre-deployed energy consumption analysis model to determine the energy consumption prediction result. The control action acquisition module is used to construct a control action state space based on the energy consumption prediction results, and to identify the control action data using a pre-deployed control action prediction model to determine the heating energy-saving control action. The energy-saving control module is used to transmit the heating energy-saving control action to the controller of the heating system, so that the controller of the heating system can perform heating energy-saving control.
[0014] This invention provides a data analysis-based heating energy-saving analysis method and system. It acquires multi-dimensional target energy consumption data related to the heating area, identifies this data using a pre-deployed energy consumption analysis model, determines energy consumption prediction results, constructs a control action state space based on these predictions, and identifies the control action data using a pre-deployed control action prediction model to determine heating energy-saving control actions. Finally, the heating energy-saving control actions are transmitted to the heating system controller, enabling the controller to perform heating energy-saving control. This method effectively captures complex nonlinear relationships and long-term time-series dependencies in the energy supply and consumption data. Compared to traditional statistical models or shallow learning methods, it significantly improves the accuracy of energy consumption prediction. Based on accurate load forecasting and combined with deep reinforcement learning algorithms, it can generate better heating strategies and avoid energy waste. Attached Figure Description
[0015] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0016] Figure 1 A flowchart illustrating a data analysis-based energy-saving analysis method for heating systems, provided as an embodiment of the present invention; Figure 2 A schematic diagram of a data analysis-based heating energy-saving analysis system provided in an embodiment of the present invention; Among them, 201-energy consumption prediction module, 202-control action acquisition module, and 203-energy saving control module.
[0017] The accompanying drawings have illustrated specific embodiments of the invention, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the invention in any way, but rather to illustrate the concept of the invention to those skilled in the art through reference to particular embodiments. Detailed Implementation
[0018] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.
[0019] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0020] like Figure 1 As shown, this embodiment of the invention provides a data analysis-based method for energy-saving heating analysis, including: S101. Obtain multi-dimensional data of target energy consumption related to the heating area, and use a pre-deployed energy consumption analysis model to identify the multi-dimensional data of target energy consumption and determine the energy consumption prediction results. Multi-dimensional data on target energy consumption related to the heating area can be some relevant data that affect or reflect energy consumption. By using deep learning models to identify these multi-dimensional data on target energy consumption, energy consumption prediction can be achieved.
[0021] This invention uses a deep learning model to effectively capture the complex nonlinear relationships and long-term time-series dependencies in energy supply and consumption data, significantly improving the accuracy of energy consumption prediction compared to traditional statistical models or shallow learning methods.
[0022] S102. Construct a control action state space based on the energy consumption prediction results, and use a pre-deployed control action prediction model to identify the control action data and determine the heating energy-saving control action. This invention, based on accurate load forecasting and combined with advanced deep reinforcement learning algorithms, can generate better heating strategies, avoid energy waste, and thus achieve energy saving.
[0023] S103. The heating energy-saving control action is transmitted to the controller of the heating system so that the controller of the heating system can perform heating energy-saving control.
[0024] This invention can process multi-dimensional and high-dimensional input data and adapt to changes in heating system characteristics, weather variations, and user behavior patterns through iterative model updates, maintaining long-term energy-saving effects. Through refined and intelligent adjustments to heating strategies, it can minimize unnecessary energy consumption while ensuring user comfort, achieving significant energy-saving goals.
[0025] In one possible implementation, multi-dimensional data on target energy consumption related to the heating area are acquired, including: By analyzing water supply temperature, return water temperature, outdoor temperature, humidity, wind speed, solar radiation intensity, building load characteristic parameters, historical user heating behavior data, historical heating supply, electricity consumption, and water consumption, multi-dimensional data on target energy consumption related to the heating area can be obtained.
[0026] Building load characteristic parameters can be related parameters such as building area, thermal insulation performance, window-to-wall ratio, etc. Historical user heat behavior data can be such as window opening frequency, indoor temperature setting, etc., or can be indoor temperature setting collected by sensors or reported by the program.
[0027] Data sampling frequency can be measured in hours. By collecting multi-dimensional data on target energy consumption at N historical time points and then constructing a data matrix, multi-dimensional data on target energy consumption related to the heating area can be obtained.
[0028] In one possible implementation, a pre-deployed energy consumption analysis model is used to identify the multi-dimensional data of the target energy consumption and determine the energy consumption prediction results, including: The target energy consumption multi-dimensional data is normalized to obtain normalized target energy consumption multi-dimensional data. The normalized multi-dimensional target energy consumption data is input into the pre-deployed energy consumption analysis model to obtain energy consumption prediction results.
[0029] In one possible implementation, the pre-deployment method of the energy consumption analysis model includes: A1. Construct an energy consumption analysis model using a convolutional neural network, and initialize the model parameters of the energy consumption analysis model to obtain multiple different individual parameters; Convolutional neural networks (CNNs) have the ability to recognize data matrices; therefore, they are preferred for constructing energy analysis models. The model parameters of a CNN typically consist of weights and thresholds, which usually have upper and lower limits. By randomly initializing the model parameters within these limits and encoding the randomly initialized parameters into vectors, individual parameters can be obtained. Repeating this process multiple times yields a variety of different individual parameters.
[0030] A2. Obtain historical energy consumption multi-dimensional data and the corresponding historical energy consumption data, and obtain the fitness of each individual parameter based on the historical energy consumption multi-dimensional data and the corresponding historical energy consumption data. Historical energy consumption multi-dimensional data refers to energy consumption data at N historical time points. The corresponding historical energy consumption data can refer to the energy consumption at the (N+1)th time point. By learning the data relationship between these two points, the energy consumption analysis model can acquire energy consumption prediction capabilities. It is worth noting that the historical energy consumption multi-dimensional data also needs to be normalized before being input into the energy consumption analysis model to reduce data complexity.
[0031] You can first obtain the cross-entropy loss function or the root mean square loss function, then add the cross-entropy loss function or the root mean square loss function to a preset constant term (such as 0.0001), and take the reciprocal to obtain the fitness.
[0032] A3. Based on the fitness of the individual parameter, perform an adaptive neighborhood search on the individual parameter to obtain the individual parameter after the adaptive neighborhood search:
[0033]
[0034]
[0035] in, Indicates the first t During the training process, the first i The parameter individual corresponds to the first... d 3D model parameters, i =1,2,…,NP, where NP represents the total number of individuals with the parameter. d =1,2,…,D, where D represents the total dimension of the model parameters in each parameter individual. Indicates the first i The parameter individual corresponding to the first adaptive neighborhood search d 3D model parameters, Represents the first random number between (0,1). Indicates the first t +1 adaptive neighborhood search factor during training. Indicates the first d The difference between the upper and lower bounds of the dimensional model parameters. Indicates the first t Adaptive neighborhood search factor during the training process. Denotes the first attenuation coefficient, and Greater than 1, Indicates the first t During the training process, the first i The neighborhood search control coefficients corresponding to each parameter individual. Indicates the first t During the training process, the first i The fitness of each individual parameter Indicates the first t The maximum fitness of all parameters for each individual during the training process. Indicates the first t The minimum fitness of all individual parameters during the training process. This represents a constant term, and is set to 0.001.
[0036] Adaptive neighborhood search is performed on the parameter individuals, which can adaptively adjust the search step size of the parameter individuals according to their fitness, so that parameter individuals with smaller fitness have a larger search range. As the algorithm progresses, all parameter individuals gradually cluster together, thereby gradually increasing the search accuracy and ensuring the convergence of the algorithm.
[0037] A4. Perform a joint search on the individual parameters after the adaptive neighborhood search to obtain the individual parameters after the joint search:
[0038]
[0039]
[0040] in, Indicates the first t During the training process, the first n The parameter individual corresponding to the first adaptive neighborhood search d 3D model parameters, Indicates the first n The parameter individual corresponding to the joint search of individuals d 3D model parameters, Represents the boundary information fusion factor. Indicates joint search factors, Indicates the first d The upper limit of the parameters of the dimensional model. Indicates the first d The lower bound of the parameters of the dimensional model. Indicates the first t During the training process, the first m The parameter individual corresponding to the first adaptive neighborhood search d 3D model parameters, Indicates the first t During the training process, the first n The parameters of the individual after the adaptive neighborhood search and the first individual m The Euclidean distance between individual parameters after adaptive neighborhood search. This indicates the preset maximum number of training iterations. This represents the preset maximum value of the boundary information fusion factor. This represents the preset minimum value of the boundary information fusion factor. The coefficient represents the joint search control coefficient, and e represents the natural constant. This represents the joint search range factor.
[0041] Performing a joint search on the individual parameters after the adaptive neighborhood search can effectively fuse the information of the individual parameters, improving the convergence accuracy of the algorithm while increasing the ability to escape local optima.
[0042] A5. Perform a positional fluctuation search on the parameter individuals after the joint search of the individuals to obtain the parameter individuals after the positional fluctuation search:
[0043] in, Indicates the first t During the training process, the first j The parameter individual corresponding to the joint search of individuals d 3D model parameters, Indicates the first j The parameter individual corresponding to the position fluctuation search is the first one. d 3D model parameters, Represented as the first t During the training process, the first j After a joint search of individuals, the parameter individuals are randomly matched with other parameter individuals corresponding to the first parameter individuals. d 3D model parameters, This indicates that a second attenuation coefficient is randomly generated according to a normal distribution N(0,1). This represents the second random number between (0,1). This represents an exponential function with the natural constant e as its base.
[0044] By performing positional fluctuation search on the parameter individuals after the joint search of the individuals, and combining the position of the parameter individuals in the solution space with the information of other parameter individuals, the ability to explore unknown solution space regions can be improved, thus enhancing the global search capability.
[0045] A6. Perform information fusion mutation search on the parameter individuals after the position fluctuation search to obtain the parameter individuals after information fusion mutation search as follows:
[0046] c =0.8*(1- t / T ) in, Indicates the first t During the training process, the first k Individual parameters after positional fluctuation search, Indicates the first k The parameters of an individual after information fusion and mutation search. c Indicates the information fusion variation factor. This represents the first randomly generated individual. This represents a second randomly generated individual.
[0047] Performing information fusion and mutation search on the individual parameters after the position fluctuation search can effectively increase the global search capability, and as the algorithm progresses, it gradually transforms into a fine search, ensuring the convergence capability of the algorithm.
[0048] A7. Determine whether the current number of training iterations is greater than or equal to the maximum number of training iterations. If so, determine the target optimal individual based on the parameter individuals after information fusion mutation search. Otherwise, return to the step of obtaining the fitness of each parameter individual. A8. Use the model parameters in the target optimal individual as the final model parameters of the energy consumption analysis model, and pre-deploy the energy consumption analysis model.
[0049] By combining the above search methods, the training effect of the algorithm can be effectively improved, thereby ensuring that the energy analysis model can make accurate predictions.
[0050] In one possible implementation, a control action state space is constructed based on the energy consumption prediction results, including: The energy consumption prediction results, current water supply temperature, current return water temperature, outdoor temperature, upper limit of pipeline pressure, lower limit of pipeline pressure, upper limit of heat source output, and lower limit of heat source output are used to construct the control action state space.
[0051] It is worth noting that, in addition to the parameters mentioned above, other parameters can also be used to construct the control action state space.
[0052] In one possible implementation, a pre-deployed control action prediction model is used to identify the control action data and determine the heating energy-saving control actions, including: The control action state space is input into a pre-deployed control action prediction model to obtain the heating energy-saving control action output by the pre-deployed control action prediction model; wherein, the heating energy-saving control action includes at least the actual output of the heat source, the supply and return water temperature of the heat exchange station secondary network, the actual output, the supply and return water temperature of the heat exchange station secondary network, the valve opening of the heating pipeline, and the actual speed of the hot water pump.
[0053] In one possible implementation, the pre-deployed control action prediction model is set up as a policy network.
[0054] In one possible implementation, after determining the heating energy-saving control action, the method further includes: Obtain the reward corresponding to the heating energy-saving control action, and construct historical training experience based on the control action state space, the heating energy-saving control action, the reward, and the control action state space after executing the heating energy-saving control action; Based on the historical training experience, the strategy network is updated in conjunction with the value network to obtain the updated strategy network. The updated strategy network is then used to determine the heating energy-saving control actions in the subsequent process.
[0055] The method for obtaining the reward includes: obtaining the difference between the predicted energy consumption and the actual energy consumption; adding the absolute value of the difference to a constant term (such as 1), and taking the reciprocal to obtain the heating reward item; the more accurate the heating, the larger the heating reward item, thus ensuring the accuracy of heating. Then, the sum of the negative energy consumption cost and the negative carbon emission is used as the energy-saving benefit item; the larger the energy-saving benefit item, the better the energy-saving effect. The reward is obtained by weighted summing of the heating reward item and the energy-saving benefit item.
[0056] Updating the policy network based on the historical training experience and incorporating the value network is a common technique in deep reinforcement learning algorithms, and will not be elaborated upon in this embodiment of the invention.
[0057] In one possible implementation, the pre-deployment method of the control action prediction model includes: pre-deploying the control action prediction model using the DQN (Deep Q-Network) algorithm.
[0058] By using deep reinforcement learning algorithms to select heating energy-saving control actions, control accuracy can be gradually improved, thereby ensuring that user energy needs are met while improving control accuracy and achieving energy-saving effects.
[0059] This invention provides a data analysis-based heating energy-saving analysis method. It acquires multi-dimensional target energy consumption data related to the heating area and uses a pre-deployed energy consumption analysis model to identify the target energy consumption data and determine the energy consumption prediction results. Then, based on the energy consumption prediction results, a control action state space is constructed, and a pre-deployed control action prediction model is used to identify the control action data and determine the heating energy-saving control actions. Finally, the heating energy-saving control actions are transmitted to the controller of the heating system, enabling the controller to perform heating energy-saving control. This method effectively captures complex nonlinear relationships and long-term time-series dependencies in the energy supply and consumption data. Compared with traditional statistical models or shallow learning methods, it significantly improves the accuracy of energy consumption prediction. Based on accurate load forecasting and combined with deep reinforcement learning algorithms, it can generate better heating strategies and avoid energy waste.
[0060] like Figure 2 As shown, this embodiment of the invention provides a heating energy-saving analysis system based on data analysis, including: an energy consumption prediction module 201, a control action acquisition module 202, and an energy-saving control module 203; The energy consumption prediction module 201 is used to acquire multi-dimensional data of target energy consumption related to the heating area, and to identify the multi-dimensional data of target energy consumption using a pre-deployed energy consumption analysis model to determine the energy consumption prediction result. The control action acquisition module 202 is used to construct a control action state space based on the energy consumption prediction results, and to identify the control action data using a pre-deployed control action prediction model to determine the heating energy-saving control action. The energy-saving control module 203 is used to transmit the heating energy-saving control action to the controller of the heating system, so that the controller of the heating system can perform heating energy-saving control.
[0061] The heating energy-saving analysis system based on data analysis provided in this embodiment of the invention can execute the above-mentioned method and technical solution. Its principle and beneficial effects are similar, and will not be repeated here.
[0062] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. It should be understood that the invention is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.
Claims
1. A data-based method for energy-saving heating analysis, characterized in that, include: Acquire multi-dimensional data on target energy consumption related to the heating area, and use a pre-deployed energy consumption analysis model to identify the multi-dimensional data on target energy consumption and determine the energy consumption prediction results; Based on the energy consumption prediction results, a control action state space is constructed, and a pre-deployed control action prediction model is used to identify the control action data to determine the heating energy-saving control action. The heating energy-saving control action is transmitted to the controller of the heating system so that the controller of the heating system can perform heating energy-saving control.
2. The heating energy-saving analysis method based on data analysis according to claim 1, characterized in that, Obtain multi-dimensional data on target energy consumption related to the heating area, including: By analyzing water supply temperature, return water temperature, outdoor temperature, humidity, wind speed, solar radiation intensity, building load characteristic parameters, historical user heating behavior data, historical heating supply, electricity consumption, and water consumption, multi-dimensional data on target energy consumption related to the heating area can be obtained.
3. The heating energy-saving analysis method based on data analysis according to claim 1, characterized in that, A pre-deployed energy consumption analysis model is used to identify the multi-dimensional data of the target energy consumption and determine the energy consumption prediction results, including: The target energy consumption multi-dimensional data is normalized to obtain normalized target energy consumption multi-dimensional data. The normalized multi-dimensional target energy consumption data is input into the pre-deployed energy consumption analysis model to obtain energy consumption prediction results.
4. The heating energy-saving analysis method based on data analysis according to claim 1, characterized in that, The pre-deployment method for the energy consumption analysis model includes: A convolutional neural network is used to construct an energy consumption analysis model, and the model parameters of the energy consumption analysis model are initialized to obtain multiple different individual parameters; Acquire historical energy consumption multi-dimensional data and the corresponding historical energy consumption data, and obtain the fitness of each individual parameter based on the historical energy consumption multi-dimensional data and the corresponding historical energy consumption data. Based on the fitness of the individual parameter, an adaptive neighborhood search is performed on the individual parameter to obtain the individual parameter after the adaptive neighborhood search. Perform a joint search on the individual parameters after the adaptive neighborhood search to obtain the individual parameters after the joint search. Perform positional fluctuation search on the parameter individuals after the joint search of the individuals to obtain the parameter individuals after the positional fluctuation search; The parameter individuals after the position fluctuation search are subjected to information fusion mutation search to obtain the parameter individuals after information fusion mutation search. Determine if the current number of training iterations is greater than or equal to the maximum number of training iterations. If so, determine the target optimal individual based on the parameter individuals after information fusion mutation search. Otherwise, return to the step of obtaining the fitness of each parameter individual. The model parameters in the target optimal individual are used as the final model parameters of the energy consumption analysis model, and the energy consumption analysis model is pre-deployed.
5. The heating energy-saving analysis method based on data analysis according to claim 1, characterized in that, Based on the energy consumption prediction results, a control action state space is constructed, including: The energy consumption prediction results, current water supply temperature, current return water temperature, outdoor temperature, upper limit of pipeline pressure, lower limit of pipeline pressure, upper limit of heat source output, and lower limit of heat source output are used to construct the control action state space.
6. The heating energy-saving analysis method based on data analysis according to claim 1, characterized in that, The control action data is identified using a pre-deployed control action prediction model to determine the heating energy-saving control actions, including: The control action state space is input into a pre-deployed control action prediction model to obtain the heating energy-saving control action output by the pre-deployed control action prediction model; wherein, the heating energy-saving control action includes at least the actual output of the heat source, the supply and return water temperatures of the secondary network of the heat exchange station, the valve opening of the heating pipeline, and the actual speed of the hot water pump.
7. The heating energy-saving analysis method based on data analysis according to claim 6, characterized in that, The pre-deployed control action prediction model is set as a policy network.
8. The heating energy-saving analysis method based on data analysis according to claim 7, characterized in that, After determining the heating energy-saving control actions, the following are also included: Obtain the reward corresponding to the heating energy-saving control action, and construct historical training experience based on the control action state space, the heating energy-saving control action, the reward, and the control action state space after executing the heating energy-saving control action; Based on the historical training experience, the strategy network is updated in conjunction with the value network to obtain the updated strategy network. The updated strategy network is then used to determine the heating energy-saving control actions in the subsequent process.
9. The heating energy-saving analysis method based on data analysis according to claim 7, characterized in that, The pre-deployment method for the control action prediction model includes: pre-deploying the control action prediction model using the DQN algorithm.
10. A heating energy-saving analysis system based on data analysis, characterized in that, include: Energy prediction module, control action acquisition module, and energy-saving control module; The energy consumption prediction module is used to acquire multi-dimensional data of target energy consumption related to the heating area, and to identify the multi-dimensional data of target energy consumption using a pre-deployed energy consumption analysis model to determine the energy consumption prediction result. The control action acquisition module is used to construct a control action state space based on the energy consumption prediction results, and to identify the control action data using a pre-deployed control action prediction model to determine the heating energy-saving control action. The energy-saving control module is used to transmit the heating energy-saving control action to the controller of the heating system, so that the controller of the heating system can perform heating energy-saving control.