Optical storage and charging station multi-strategy operation method, device and equipment based on deep learning

Abstract

The invention discloses an optical storage and charging station multi-strategy operation method, device and equipment based on deep learning. The method comprises the following steps: acquiring multi-source heterogeneous operation data of an optical storage and charging integrated energy station; based on the improved space-time heterogeneous graph fusion model, performing data fusion on the operation data by using an improved gating graph convolutional network to obtain fused data; performing feature extraction operation on the fused data to obtain a high-dimensional feature vector; carrying out dimensionality reduction and key information extraction on the high-dimensional feature vector through an improved sparse auto-encoder to generate a low-dimensional representation, and carrying out global search on the low-dimensional representation through an improved hybrid evolutionary algorithm to generate an optimal operation strategy; and applying the optimal operation strategy to the operation of the optical storage and charging integrated energy station in real time. The operation management efficiency and the operation management effect of the optical storage and charging integrated energy station can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a multi-strategy operation method, device and equipment for a photovoltaic storage and charging station based on deep learning. Background Art

[0002] In the operation scenario of integrated photovoltaic, storage and charging energy stations, existing technologies generally face many challenges. On the one hand, when processing multi-source heterogeneous data, existing systems often rely on traditional data analysis methods, making it difficult to accurately capture the complex coupling relationship between photovoltaic panels, energy storage batteries and charging pile data. For example, under different climatic conditions and peak and valley periods of electricity consumption, the coordinated operation of photovoltaic power generation, energy storage allocation and charging demand is extremely complex, and existing technologies cannot fully explore the deep correlation hidden behind these data, resulting in limited comprehensive energy utilization efficiency. On the other hand, the current operation strategies are mostly based on experience or simple historical data statistics, lacking the ability to adaptively adjust to real-time dynamic changes. When the power price of the power grid fluctuates and the photovoltaic power generation changes significantly due to the weather, it is difficult to quickly make the optimal charging pile pricing and energy storage charging and discharging decisions, resulting in high operating costs and difficult to achieve the best user experience. This lack of dynamic and targeted operation strategy is difficult to adapt to the complex and changeable operating environment of the integrated photovoltaic, storage and charging energy station, resulting in low energy utilization efficiency and high operating costs.

[0003] In short, the existing operation mode of integrated photovoltaic, storage and charging energy stations is difficult to effectively adapt to the complex and changeable operating environment of integrated photovoltaic, storage and charging energy stations, resulting in low operation and management efficiency and unsatisfactory operation and management results. Summary of the invention

[0004] The embodiments of the present invention provide a multi-strategy operation method, device and equipment for a photovoltaic storage and charging station based on deep learning, which can effectively improve the operation and management efficiency and operation and management effect of the photovoltaic storage and charging integrated energy station.

[0005] An embodiment of the present invention provides a multi-strategy operation method of a photovoltaic storage and charging station based on deep learning, comprising: Acquire the multi-source heterogeneous operation data of the photovoltaic storage and charging integrated energy station, which includes the status data of photovoltaic panels, energy storage batteries and the usage status data of charging piles; Based on an improved spatiotemporal heterogeneous graph fusion model and using an improved gated graph convolutional network, data fusion is performed on the operation data to obtain fused data; Perform feature extraction on the fused data to obtain a high-dimensional feature vector that comprehensively reflects the complex dependencies between multi-source heterogeneous operational data; The high-dimensional feature vector is reduced in dimension and key information is extracted by an improved sparse autoencoder to generate a low-dimensional representation, and the low-dimensional representation is globally searched by an improved hybrid evolutionary algorithm to generate an optimal operation strategy; the optimal operation strategy includes a charging pile pricing strategy, an energy storage battery charging and discharging strategy, and a user discount push strategy; The optimal operation strategy is applied in real time to the operation of the photovoltaic, storage and charging integrated energy station; wherein, the pricing of the charging piles of the photovoltaic, storage and charging integrated energy station is dynamically adjusted according to the charging pile pricing strategy and through smart meters, discount information is pushed to user terminals according to the user preferential push strategy, and the charging and discharging strategy of the energy storage battery of the photovoltaic, storage and charging integrated energy station is optimized according to the energy storage battery charging and discharging strategy and through an energy storage management system.

[0006] As an improvement of the above solution, the operation data is fused based on the improved spatiotemporal heterogeneous graph fusion model and using an improved gated graph convolutional network to obtain fused data, including: The operation data is normalized using a normalization algorithm that combines logarithmic transformation and linear scaling, thereby mapping the operation data to the adaptation interval to obtain the preprocessed operation data; According to the preprocessed operation data, the photovoltaic panels, energy storage batteries and charging piles of the photovoltaic storage and charging integrated energy station are abstracted into nodes of a spatiotemporal heterogeneous graph and their attributes are expanded. At the same time, the time edges and space edges of the spatiotemporal heterogeneous graph are defined, and the weights of each edge of the spatiotemporal heterogeneous graph are calculated to reflect the spatiotemporal correlation between the nodes, and the spatiotemporal heterogeneous graph is obtained by composing the graph; By using an improved gated graph convolutional network, a multi-layer network is operated layer by layer on the constructed spatiotemporal heterogeneous graph to aggregate the local features of the nodes, thereby obtaining fused data.

[0007] As an improvement to the above solution, the feature extraction operation is performed on the fused data to obtain a high-dimensional feature vector that comprehensively reflects the complex dependency relationship between multi-source heterogeneous operation data, including: Reshape the fused data and build it into a data cube structure according to different data sources and time series. At the same time, dynamically divide the windows according to the data fluctuation conditions to generate adaptive dynamic window data. Based on the obtained dynamic window data, the comprehensive importance of each feature is evaluated in each window from the information entropy and feature correlation level, and the important features are screened out to form a screened feature subset; For the filtered feature subset, a deep residual attention network is built, and the residual connection is used to solve the gradient problem. The attention mechanism is used to strengthen the expression of key features. After processing, global average pooling is performed to obtain a high-dimensional feature vector that comprehensively reflects the complex dependencies between multi-source heterogeneous operation data.

[0008] As an improvement of the above scheme, the high-dimensional feature vector is reduced in dimension and key information is extracted by an improved sparse autoencoder to generate a low-dimensional representation, and the low-dimensional representation is globally searched by an improved hybrid evolutionary algorithm to generate an optimal operation strategy; the optimal operation strategy includes a charging pile pricing strategy, an energy storage battery charging and discharging strategy, and a user discount push strategy, including: The high-dimensional feature vector is first preprocessed based on data distribution characteristics, and a low-dimensional representation after dimensionality reduction is generated by constructing an information topology map and encoding it using a unique probability activation function, while introducing a sparse constraint mechanism based on information gain rate for training; Based on the low-dimensional representation, an initial population is generated using chaotic mapping, and individuals in the population correspond to different combinations of operation strategies; Construct a multi-objective fitness function that comprehensively considers the total revenue of the energy station, user satisfaction and energy utilization rate, and evaluate the fitness of the initial population; Use chaos search to generate disturbance vectors in the neighborhood of the initial population individuals, and update the population by comparing the fitness of the new individuals with the original individuals to achieve global exploration; Based on the population after chaotic search, a multi-objective game mechanism is introduced. Individuals generate new individuals in the neighborhood, play games based on different operating target benefits, adjust their own strategy parameters, and promote population evolution. After multiple iterations of chaotic search and multi-objective game, when the termination condition is met, the strategy corresponding to the individual with the highest fitness in the population is selected to generate the optimal operation strategy including charging pile pricing, energy storage battery charging and discharging, and user discount push.

[0009] Another embodiment of the present invention provides a multi-strategy operation device for a photovoltaic storage and charging station based on deep learning, including: The data acquisition module is used to obtain the multi-source heterogeneous operation data of the photovoltaic storage and charging integrated energy station, and the operation data includes the status data of photovoltaic panels, energy storage batteries and the usage status data of charging piles; A data fusion module, used for fusing the operation data based on an improved spatiotemporal heterogeneous graph fusion model and using an improved gated graph convolutional network to obtain fused data; The feature extraction module is used to perform feature extraction operations on the fused data to obtain a high-dimensional feature vector that comprehensively reflects the complex dependencies between multi-source heterogeneous operational data; A strategy generation module, which is used to reduce the dimension and extract key information of the high-dimensional feature vector by an improved sparse autoencoder to generate a low-dimensional representation, and to perform a global search on the low-dimensional representation by an improved hybrid evolutionary algorithm to generate an optimal operation strategy; the optimal operation strategy includes a charging pile pricing strategy, an energy storage battery charging and discharging strategy, and a user discount push strategy; A strategy application module is used to apply the optimal operation strategy to the operation of the photovoltaic, storage and charging integrated energy station in real time; wherein, the pricing of the charging piles of the photovoltaic, storage and charging integrated energy station is dynamically adjusted according to the charging pile pricing strategy and through smart meters, discount information is pushed to user terminals according to the user preferential push strategy, and the charging and discharging strategy of the energy storage battery of the photovoltaic, storage and charging integrated energy station is optimized according to the energy storage battery charging and discharging strategy and through an energy storage management system.

[0010] As an improvement of the above solution, the data fusion module is specifically used for: The operation data is normalized using a normalization algorithm that combines logarithmic transformation and linear scaling, thereby mapping the operation data to the adaptation interval to obtain the preprocessed operation data; According to the preprocessed operation data, the photovoltaic panels, energy storage batteries and charging piles of the photovoltaic storage and charging integrated energy station are abstracted into nodes of a spatiotemporal heterogeneous graph and their attributes are expanded. At the same time, the time edges and space edges of the spatiotemporal heterogeneous graph are defined, and the weights of each edge of the spatiotemporal heterogeneous graph are calculated to reflect the spatiotemporal correlation between the nodes, and the spatiotemporal heterogeneous graph is obtained by composing the graph; By using an improved gated graph convolutional network, a multi-layer network is operated layer by layer on the constructed spatiotemporal heterogeneous graph to aggregate the local features of the nodes, thereby obtaining fused data.

[0011] As an improvement of the above solution, the feature extraction module is specifically used for: Reshape the fused data and build it into a data cube structure according to different data sources and time series. At the same time, dynamically divide the windows according to the data fluctuation conditions to generate adaptive dynamic window data. Based on the obtained dynamic window data, the comprehensive importance of each feature is evaluated in each window from the information entropy and feature correlation level, and the important features are screened out to form a screened feature subset; For the filtered feature subset, a deep residual attention network is built, and the residual connection is used to solve the gradient problem. The attention mechanism is used to strengthen the expression of key features. After processing, global average pooling is performed to obtain a high-dimensional feature vector that comprehensively reflects the complex dependencies between multi-source heterogeneous operation data.

[0012] As an improvement of the above solution, the strategy generation module is specifically used for: The high-dimensional feature vector is first preprocessed based on data distribution characteristics, and a low-dimensional representation after dimensionality reduction is generated by constructing an information topology map and encoding it using a unique probability activation function, while introducing a sparse constraint mechanism based on information gain rate for training; Based on the low-dimensional representation, an initial population is generated using chaotic mapping, and individuals in the population correspond to different combinations of operation strategies; Construct a multi-objective fitness function that comprehensively considers the total revenue of the energy station, user satisfaction and energy utilization rate, and evaluate the fitness of the initial population; Use chaos search to generate disturbance vectors in the neighborhood of the initial population individuals, and update the population by comparing the fitness of the new individuals with the original individuals to achieve global exploration; Based on the population after chaotic search, a multi-objective game mechanism is introduced. Individuals generate new individuals in the neighborhood, play games based on different operating target benefits, adjust their own strategy parameters, and promote population evolution. After multiple iterations of chaotic search and multi-objective game, when the termination condition is met, the strategy corresponding to the individual with the highest fitness in the population is selected to generate the optimal operation strategy including charging pile pricing, energy storage battery charging and discharging, and user discount push.

[0013] Another embodiment of the present invention provides a multi-strategy operation device for a photovoltaic storage and charging station based on deep learning, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the multi-strategy operation method for a photovoltaic storage and charging station based on deep learning described in the above-mentioned embodiment of the invention.

[0014] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: First, we obtain multi-source heterogeneous operation data covering the status of photovoltaic panels, energy storage batteries and charging piles to provide a comprehensive information basis for subsequent analysis. We use an improved spatiotemporal heterogeneous graph fusion model and a gated graph convolutional network for data fusion. The improved model can accurately capture the complex associations between different data sources in time and space dimensions, overcome the defect that traditional methods cannot adapt to dynamic changes, and enable the fused data to more accurately reflect the actual operation status of the energy station. We extract features from the fused data to obtain high-dimensional feature vectors, fully explore the complex dependencies between multi-source heterogeneous data, and provide a key basis for subsequent strategy formulation. We use an improved sparse autoencoder to reduce the dimension of high-dimensional feature vectors and extract key information to avoid data redundancy and improve computational efficiency. We use an improved hybrid evolutionary algorithm to perform a global search on low-dimensional representations, and can quickly find the optimal operation strategy in a complex solution space. This strategy covers charging pile pricing, energy storage battery charging and discharging, and user discount push, and has strong pertinence and dynamic adaptability. Finally, we apply the optimal strategy to energy station operations in real time, and achieve precise control with the help of smart meters, user terminals and energy storage management systems. The embodiments of the present invention use innovative data fusion, feature extraction and strategy generation technologies to deeply mine the complex dependencies between multi-source heterogeneous operation data, accurately and in real time formulate and apply the optimal operation strategy, and significantly improve the energy utilization efficiency and intelligent operation level of the integrated photovoltaic storage and charging energy station. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flowchart of a multi-strategy operation method of a photovoltaic storage and charging station based on deep learning provided by an embodiment of the present invention; Figure 2 It is a structural schematic diagram of a multi-strategy operation device for a photovoltaic storage and charging station based on deep learning provided by an embodiment of the present invention; Figure 3 It is a structural schematic diagram of a multi-strategy operation device of a photovoltaic storage and charging station based on deep learning provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] See also Figure 1 , is a flowchart of a multi-strategy operation method of a photovoltaic storage and charging station based on deep learning provided by an embodiment of the present invention. The multi-strategy operation method of a photovoltaic storage and charging station based on deep learning comprises the following steps: S10, obtaining multi-source heterogeneous operation data of the photovoltaic storage and charging integrated energy station, wherein the operation data includes status data of photovoltaic panels, energy storage batteries and usage status data of charging piles; S11, based on the improved spatiotemporal heterogeneous graph fusion model and using the improved gated graph convolutional network, performing data fusion on the operation data to obtain fused data; S12, performing feature extraction on the fused data to obtain a high-dimensional feature vector that comprehensively reflects the complex dependency relationship between multi-source heterogeneous operation data; S13, reducing the dimension and extracting key information of the high-dimensional feature vector by an improved sparse autoencoder to generate a low-dimensional representation, and performing a global search on the low-dimensional representation by an improved hybrid evolutionary algorithm to generate an optimal operation strategy; the optimal operation strategy includes a charging pile pricing strategy, an energy storage battery charging and discharging strategy, and a user discount push strategy; S14, applying the optimal operation strategy to the operation of the photovoltaic, storage and charging integrated energy station in real time; wherein, the pricing of the charging piles of the photovoltaic, storage and charging integrated energy station is dynamically adjusted according to the charging pile pricing strategy and through smart meters, discount information is pushed to user terminals according to the user preferential push strategy, and the charging and discharging strategy of the energy storage battery of the photovoltaic, storage and charging integrated energy station is optimized according to the energy storage battery charging and discharging strategy and through an energy storage management system.

[0018] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: First, we obtain multi-source heterogeneous operation data covering the status of photovoltaic panels, energy storage batteries and charging piles to provide a comprehensive information basis for subsequent analysis. We use an improved spatiotemporal heterogeneous graph fusion model and a gated graph convolutional network for data fusion. The improved model can accurately capture the complex associations between different data sources in time and space dimensions, overcome the defect that traditional methods cannot adapt to dynamic changes, and enable the fused data to more accurately reflect the actual operation status of the energy station. We extract features from the fused data to obtain high-dimensional feature vectors, fully explore the complex dependencies between multi-source heterogeneous data, and provide a key basis for subsequent strategy formulation. We use an improved sparse autoencoder to reduce the dimension of high-dimensional feature vectors and extract key information to avoid data redundancy and improve computational efficiency. We use an improved hybrid evolutionary algorithm to perform a global search on low-dimensional representations, and can quickly find the optimal operation strategy in a complex solution space. This strategy covers charging pile pricing, energy storage battery charging and discharging, and user discount push, and has strong pertinence and dynamic adaptability. Finally, we apply the optimal strategy to energy station operations in real time, and achieve precise control with the help of smart meters, user terminals and energy storage management systems. The embodiments of the present invention use innovative data fusion, feature extraction and strategy generation technologies to deeply mine the complex dependencies between multi-source heterogeneous operation data, accurately and in real time formulate and apply the optimal operation strategy, and significantly improve the energy utilization efficiency and intelligent operation level of the integrated photovoltaic storage and charging energy station.

[0019] As one example of the above scheme, the acquisition of multi-source heterogeneous operation data of the photovoltaic storage and charging integrated energy station, wherein the operation data includes the status data of photovoltaic panels, energy storage batteries and the usage status data of charging piles, includes the following steps: Step 1: Determine the data type and collection point: Identify the type of operational data that needs to be obtained, namely the status data of photovoltaic panels and energy storage batteries and the usage status data of charging piles. At the same time, determine the specific collection point for each data type.

[0020] Photovoltaic panel status data collection point: In the photovoltaic panel array, sensors are installed for each photovoltaic panel group (for example, every 10 photovoltaic panels are a group) to collect data. Current sensors and voltage sensors are installed at the output end of the photovoltaic panel to obtain the power generation current and voltage data; light intensity sensors and temperature sensors are installed on the surface of the photovoltaic panel to measure the light intensity and the surface temperature of the photovoltaic panel respectively.

[0021] Energy storage battery status data collection point: In the energy storage battery pack, each battery module is equipped with a sensor. Voltage sensors and current sensors are installed at the positive and negative poles of the battery module to monitor the battery's charge and discharge voltage and current; temperature sensors are installed inside the battery module to monitor the battery's temperature changes in real time; at the same time, the battery's remaining power (SOC), health status (SOH) and other data are obtained through the battery management system (BMS).

[0022] Charging pile usage status data collection point: Install corresponding sensors and monitoring equipment on each charging pile. Install current sensors and voltage sensors at the charging interface of the charging pile to record the current and voltage during the charging process; in the control unit of the charging pile, obtain data such as charging time, charging power, user information (such as user ID, charging start and end time), etc.

[0023] Step 2: Select appropriate data acquisition equipment: Select appropriate data acquisition equipment based on the determined data type and collection points to ensure the accuracy and stability of data collection.

[0024] Sensors: For current and voltage sensors, choose high-precision, wide-range Hall effect sensors that can accurately measure data in different current and voltage ranges. Light intensity sensors use photodiode type sensors, which have the characteristics of fast response and high sensitivity. Temperature sensors use thermocouples or thermistor sensors, depending on the different measurement ranges and accuracy requirements.

[0025] Data collector: An industrial-grade data collector is used, which has multiple analog input and digital input channels and can collect data from multiple sensors at the same time. The data collector supports multiple communication protocols, such as Modbus, CAN bus, etc., which facilitates communication with sensors and subsequent data processing systems.

[0026] Communication equipment: Select appropriate communication equipment based on the distance between the collection point and the data processing center and the communication environment. For collection points that are close, wired Ethernet communication can be used; for collection points that are far away or dispersed, wireless communication methods such as GPRS, LoRa, etc. are used.

[0027] Step 3: Set the data collection frequency: Set a reasonable data collection frequency based on the changing characteristics of different data types and actual operational needs.

[0028] Photovoltaic panel status data: Since the light intensity and the temperature of the photovoltaic panel change rapidly with time and weather conditions during the day, the collection frequency of light intensity and temperature data is set to once a minute to capture these changes. The collection frequency of power generation current and voltage data is set to once every 5 minutes because they change relatively slowly.

[0029] Energy storage battery status data: The battery's charge and discharge current and voltage will change significantly during the charge and discharge process, so the collection frequency is set to once every 2 minutes. The battery's temperature, remaining power, and health status change relatively slowly, so the collection frequency is set to once every 10 minutes.

[0030] Charging pile usage status data: Charging current and voltage will change in real time during the charging process, and the collection frequency is set to once per second to accurately record the charging process. Charging time, charging power and user information are collected at the beginning and end of charging, and do not need to be collected in real time.

[0031] Step 4: Data transmission and storage: The collected data is transmitted to the data processing center through communication equipment and stored. Among them, data transmission: the data collector packages the collected data according to the set communication protocol and transmits it to the data processing center through wired or wireless communication equipment. During the transmission process, data encryption and verification technology are used to ensure the security and integrity of the data. Data storage: After the data processing center receives the data, it stores it in the database. You can choose a relational database (such as MySQL) or a non-relational database (such as MongoDB) according to the structure of the data and query requirements. At the same time, in order to facilitate subsequent data processing and analysis, the stored data is classified and labeled, for example, by data type, collection time, device number, etc.

[0032] For ease of understanding, let's take the following example: Assume that a photovoltaic, storage and charging integrated energy station has 100 photovoltaic panels, divided into 10 photovoltaic panel groups; 5 groups of energy storage batteries, each group contains 10 battery modules; and 20 charging piles. During the operation of a certain day, data collection is carried out through the above steps.

[0033] Photovoltaic panel data: At 10 a.m., the light intensity sensor of the third group of photovoltaic panels collected a light intensity of 8000 lux, the temperature sensor collected a photovoltaic panel surface temperature of 30°C, the current sensor collected a power generation current of 5A, and the voltage sensor collected a power generation voltage of 220V. These data are collected once a minute and transmitted to the data processing center through wireless communication equipment and stored in the database.

[0034] Energy storage battery data: At 2 pm, the voltage sensor of the 5th battery module of the 2nd group of energy storage batteries collected a battery voltage of 3.7V, the current sensor collected a charging current of 2A, and the temperature sensor collected a battery temperature of 25°C. The battery management system sent that the remaining power of the battery module was 80% and the health status was good. These data are collected every 2 minutes (current, voltage) or every 10 minutes (temperature, remaining power, health status) and transmitted to the data processing center for storage.

[0035] Charging pile data: At 7pm, charging pile No. 15 began to charge an electric car. During the charging process, the current sensor collected the charging current once a second, and the voltage sensor collected the charging voltage once a second. After charging, the control unit of the charging pile recorded the charging time as 1 hour, the charging power as 20 degrees, and the user ID as 12345. After these data were transmitted to the data processing center, they were stored in the database for subsequent data analysis and operational decision-making.

[0036] As one example of the above solution, the operation data is fused based on the improved spatiotemporal heterogeneous graph fusion model and using the improved gated graph convolutional network to obtain fused data, including: The operation data is normalized using a normalization algorithm that combines logarithmic transformation and linear scaling, thereby mapping the operation data to the adaptation interval to obtain the preprocessed operation data; According to the preprocessed operation data, the photovoltaic panels, energy storage batteries and charging piles of the photovoltaic storage and charging integrated energy station are abstracted into nodes of a spatiotemporal heterogeneous graph and their attributes are expanded. At the same time, the time edges and space edges of the spatiotemporal heterogeneous graph are defined, and the weights of each edge of the spatiotemporal heterogeneous graph are calculated to reflect the spatiotemporal correlation between the nodes, and the spatiotemporal heterogeneous graph is obtained by composing the graph; By using an improved gated graph convolutional network, a multi-layer network is operated layer by layer on the constructed spatiotemporal heterogeneous graph to aggregate the local features of the nodes, thereby obtaining fused data.

[0037] In this embodiment, firstly, a unique normalization algorithm integrating logarithmic transformation and linear scaling is used to accurately adjust the scale of the operation data and map it to the adaptation interval, which effectively solves the problem that multi-source heterogeneous data is difficult to handle due to scale differences, and lays a good foundation for subsequent operations. Then, the equipment of the integrated energy station of light storage and charging is abstracted as nodes of the spatiotemporal heterogeneous graph and the attributes are expanded. At the same time, the time edge and space edge are carefully defined and the weights are calculated, the spatiotemporal correlation between nodes is fully explored, and a spatiotemporal heterogeneous graph that can fully reflect the operation logic of the energy station is constructed. Finally, the spatiotemporal heterogeneous graph is multi-layered by means of an improved gated graph convolutional network to achieve the aggregation of local features of nodes, thereby obtaining highly fused fused data that can reflect the complex dependencies between data. In summary, this embodiment can not only effectively integrate multi-source heterogeneous operation data, but also deeply explore the spatiotemporal correlation behind the data, provide high-quality data support for subsequent feature extraction and operation strategy formulation, and improve the management and operation level of the integrated energy station of light storage and charging.

[0038] Specifically, the working process of this embodiment is as follows: (1) Data preprocessing and normalization: The multi-source heterogeneous operation data obtained in step S10 has different data formats, dimensions and value ranges due to different sources. Suppose the original data is ,in Represents data type is the photovoltaic panel data, For energy storage battery data, For charging pile data, Indicates data samples. A normalization method that combines logarithmic transformation and linear scaling is used, and the formula is as follows: ; in, According to the data type The offset constant is set to prevent negative numbers or zero from appearing in logarithmic operations; Is the data type The maximum value of plus a small constant (such as 0.01) is used for scale adjustment of logarithmic transformation; is the scaling factor, is the offset, which is set according to the subsequent model requirements and is used to map the normalized data to a suitable interval. For example, for the light intensity data of photovoltaic panels, this formula is used to map its value range to the [0,1] interval for subsequent processing.

[0039] (2) Improved spatiotemporal heterogeneous graph construction: Node definition and attribute expansion: Photovoltaic panels, energy storage batteries, and charging piles are abstracted as nodes of a spatiotemporal heterogeneous graph. Each node has not only conventional attributes, but also unique dynamic attributes. For example, in addition to rated power and conversion efficiency, photovoltaic panel nodes also add future light intensity trend values ​​based on time series prediction. , and the occlusion coefficient considering the surrounding occlusion , Indicates the time step. The energy storage battery node adds a battery health index calculated based on the aging model. , and the fluctuation coefficient of the power interaction with the grid The charging pile node contains the charging demand probability based on user behavior analysis. , and the competitive coefficient of the competition relationship with surrounding charging piles .

[0040] Edge relationship modeling and weight calculation: Edges are divided into time edges and space edges. Time edges connect data of the same node at different time steps, reflecting the characteristics of time series; space edges connect different types of nodes, reflecting spatial associations.

[0041] Time edge weight ,in represents different time steps, The calculation formula is as follows: ; in, and They are The attributes are in the time step and The value of and The time steps are and The mean of all attributes, is the number of attributes, is the time decay parameter, which controls the influence of time on edge weights. is a very small constant to prevent the denominator from being zero. This formula calculates the covariance of data at different time steps and combines it with an exponential decay function to reflect the characteristic that the closer the time, the stronger the correlation.

[0042] Spatial edge weights ( Represents different types of nodes) using the information gain and mutual information based method. First, calculate the node Mutual Information Between Attributes , combined with information gain , the formula is: ; in, is the maximum value of the mutual information between all possible node pairs and is used to normalize the weights. Measures the degree of dependence between two node attributes, information gain It indicates the degree of reduction of uncertainty about the attribute of another node when the attribute of one node is known, and in this way it more accurately reflects the relationship between spatial nodes.

[0043] (3) Improved Gated Graph Convolutional Network (GGCN) Fusion: Construct a multi-layer improved gated graph convolutional network for data fusion. In each layer of GGCN, the feature update formula of the node is as follows: ; in, Representation Node In the The feature vector of the layer, Is a node The set of neighbor nodes of Is a node and The spatial edge weights between is the gating vector, which is used to control the inflow and outflow of information. The calculation formula is: ; here, It is a learnable weight vector used to adjust the influence of neighbor node information on the current node; is the bias term; It is an improved activation function, such as the LeakyReLU function that dynamically adjusts the slope according to the data distribution; It is an enhancement factor that is dynamically adjusted according to the importance of the node and the current network status, and is used to highlight key node information. Through layer-by-layer calculation of multi-layer GGCN, the local features of the nodes are gradually aggregated to obtain the fused data.

[0044] In order to facilitate the understanding of this embodiment, the following example is given here: For example, in a photovoltaic storage and charging integrated energy station located in an industrial park, the power generation data of photovoltaic panels, the charging and discharging data of energy storage batteries, and the usage data of charging piles constitute multi-source heterogeneous operation data. Since the dimensions and value ranges of these data vary greatly, for example, the power generation power of photovoltaic panels may be between tens of kilowatts and hundreds of kilowatts, and the usage time of charging piles may be between tens of minutes and hours, these data are uniformly mapped to a suitable interval by integrating the normalization algorithm of logarithmic transformation and linear scaling, which is convenient for subsequent processing. Then, the photovoltaic panels, energy storage batteries and charging piles are abstracted as nodes of a spatiotemporal heterogeneous graph, and the photovoltaic panel nodes are expanded with attributes such as light intensity and temperature in different time periods, the energy storage battery nodes are expanded with attributes such as remaining power and charging and discharging efficiency, and the charging pile nodes are expanded with attributes such as charging power and the number of waiting vehicles. At the same time, time edges are defined to reflect the change relationship of each device data over time, such as the change relationship of photovoltaic panel power generation at different times; spatial edges are defined to reflect the spatial relationship between devices, such as the energy transmission relationship between photovoltaic panels and energy storage batteries. After calculating the weights of each edge, a spatiotemporal heterogeneous graph is constructed. Finally, the improved gated graph convolutional network is used to operate on the graph, aggregate the local features of each device node, and obtain the fused data. For example, the fused data can clearly reflect the complex spatiotemporal correlations such as the increase in photovoltaic panel power generation, the start of charging of energy storage batteries, and the increase in demand for charging piles in the afternoon when the sun is sufficient, providing a data basis for the operation and management of energy stations.

[0045] As one example of the above solution, the feature extraction operation is performed on the fused data to obtain a high-dimensional feature vector that comprehensively reflects the complex dependency relationship between multi-source heterogeneous operation data, including: Reshape the fused data and build it into a data cube structure according to different data sources and time series. At the same time, dynamically divide the windows according to the data fluctuation conditions to generate adaptive dynamic window data. Based on the obtained dynamic window data, the comprehensive importance of each feature is evaluated in each window from the information entropy and feature correlation level, and the important features are screened out to form a screened feature subset; For the filtered feature subset, a deep residual attention network is built, and the residual connection is used to solve the gradient problem. The attention mechanism is used to strengthen the expression of key features. After processing, global average pooling is performed to obtain a high-dimensional feature vector that comprehensively reflects the complex dependencies between multi-source heterogeneous operation data.

[0046] In this embodiment, first, the fused data is reshaped to construct a unique data cube structure, with different data sources and time series as dimensions, breaking through the traditional data organization form, so that the multi-source attributes and time dynamics of the data can be intuitively presented. At the same time, the dynamic window division mechanism is very innovative, abandoning the fixed window mode, adjusting the window size in real time according to the data fluctuation, ensuring that the generated dynamic window data is closely fitted to the actual data, and effectively capturing the data features under different degrees of fluctuation. Then, the importance of features is evaluated from the dual dimensions of information entropy and feature correlation. This multivariate evaluation method is more comprehensive and scientific than single indicator screening, and can deeply mine the features that play a key role in reflecting the complex dependencies of data, and then screen out important feature subsets. Then, a deep residual attention network is built, and residual connections are used to overcome the gradient problem of deep networks, so as to realize the learning of deep features, focus on key features with the help of attention mechanisms, and highlight important information. Finally, the processed features are integrated through global average pooling to form a high-dimensional feature vector. In summary, this embodiment can not only efficiently process multi-source heterogeneous data and deeply mine the complex connections between data, but also greatly improve the accuracy and effectiveness of feature extraction through innovative window division, multi-dimensional evaluation and deep network structure, providing high-quality data support for subsequent dimensionality reduction and strategy generation.

[0047] Specifically, the working process of this embodiment is as follows: (1) Data structure reshaping and dynamic window division: Get the fused data Afterwards, due to its multi-source heterogeneous characteristics, the dependencies within the data vary in different time scales and spatial dimensions. In order to capture these complex relationships more accurately, the data is first restructured. Arrange data in three dimensions according to different data sources (photovoltaic panels, energy storage batteries, charging piles) and time series to form a data cube , whose dimensions are the number of samples , number of features and time step Next, we introduce a dynamic window partitioning mechanism. Traditional window partitioning methods often fix the window size and are difficult to adapt to the dynamic changes of data. Here, we dynamically adjust the window size according to the fluctuation of the data. Define a fluctuation index , for each time step , calculate the before and after The sum of the standard deviations of the data within time steps is used as a volatility indicator: ; Where std represents the standard deviation calculation function. The size of the window is determined dynamically : ; in is the base window size, is the adjustment factor, In this way, a larger window is used in areas with large data fluctuations to capture a wider range of dependencies, and a smaller window is used in areas with stable data to improve computational efficiency.

[0048] (2) Feature screening based on information entropy and correlation: In each dynamic window, the data is screened for features. Traditional feature screening methods are mostly based on a single statistical indicator, which is difficult to fully reflect the complex relationship between multi-source heterogeneous data. Here, we combine information entropy and the correlation between features for screening. For each feature in the window , calculate its information entropy : ; in It is a feature In the The information entropy reflects the uncertainty of the feature. The larger the information entropy, the richer the information contained in the feature. At the same time, the mutual information between features is calculated. To measure the correlation between features: ; middle It is a feature In the Samples and features In the Based on information entropy and mutual information, a comprehensive score is defined To evaluate the importance of each feature: ; in and is the weight coefficient, and Based on the comprehensive score Sort the features and select the ones with higher scores. features to form a filtered feature subset .

[0049] (3) Feature extraction based on deep residual attention network: In order to select the feature subset The high-dimensional feature vector that comprehensively reflects the complex dependencies of multi-source heterogeneous operation data is extracted to construct a deep residual attention network (DRAN). DRAN is composed of multiple residual attention blocks (RABs) stacked together. Each RAB contains residual connections and attention mechanisms.

[0050] Residual connection: for input features , after the convolutional layer And the activation function After processing , and then through the convolutional layer get Through residual connection, the output , which can alleviate the gradient vanishing problem and enable the network to learn deeper features. Attention mechanism: Based on the residual connection, the attention mechanism is introduced to enhance the expression of important features. For the input feature , calculate the attention weight : ; in and is a learnable weight matrix. With input features Multiply element by element to get the features enhanced by attention ,in After multiple RABs are processed, the output of the last RAB is globally averaged and pooled to obtain the final high-dimensional feature vector. This high-dimensional feature vector comprehensively reflects the complex dependencies between multi-source heterogeneous operational data and can be used for subsequent tasks such as dimensionality reduction and strategy generation.

[0051] In order to facilitate the understanding of this embodiment, the following example is given here: For example, in a photovoltaic storage and charging integrated energy station located in an industrial park, the fused data covers multi-source information such as the power generation efficiency of photovoltaic panels at different times, the charge state of energy storage batteries, and the busyness of charging piles. After these data are constructed into a data cube structure, with time as the clue, photovoltaic panels, energy storage batteries, charging piles and other equipment as the data source dimension, the internal hierarchy of the data is clearly presented. When the industrial park ushers in the peak production period, the demand for electricity surges, and the power generation efficiency of photovoltaic panels may fluctuate greatly. At this time, the dynamic window division mechanism will quickly adjust the window size according to the data fluctuation situation, and expand the window when the fluctuation is large, so as to capture more information related to changes in energy supply and demand. In each dynamic window, from the perspective of information entropy, if the information entropy of the charge state of the energy storage battery at a certain moment is high, it means that the uncertainty of the feature is large, which may have an important impact on the stable power supply of the energy station; from the perspective of feature correlation, it will be found that there is a certain correlation between the frequency of use of charging piles and the power generation efficiency of photovoltaic panels. Based on the evaluation of these two dimensions, important features such as photovoltaic panel power generation efficiency, energy storage battery charge state, and charging pile use frequency are screened out to form a feature subset. Subsequently, these feature subsets are input into the deep residual attention network. The residual connection helps the network learn more complex feature relationships between the various parts of the energy station, and the attention mechanism focuses on key features such as the sudden drop in the power generation efficiency of photovoltaic panels, and strengthens their expression. Finally, after global average pooling, a high-dimensional feature vector is obtained, which comprehensively reflects the complex dependency relationship between photovoltaic panels, energy storage batteries and charging piles during the peak production period of the industrial park. For example, when the photovoltaic panels are insufficient, how the energy storage batteries supplement the power supply, how the use of charging piles will change, etc., provides a core basis for the energy station to formulate a scientific and reasonable operation strategy.

[0052] As one example of the above solution, the high-dimensional feature vector is reduced in dimension and key information is extracted by an improved sparse autoencoder to generate a low-dimensional representation, and the low-dimensional representation is globally searched by an improved hybrid evolutionary algorithm to generate an optimal operation strategy; the optimal operation strategy includes a charging pile pricing strategy, an energy storage battery charging and discharging strategy, and a user discount push strategy, including: The high-dimensional feature vector is first preprocessed based on data distribution characteristics, and a low-dimensional representation after dimensionality reduction is generated by constructing an information topology map and encoding it using a unique probability activation function, while introducing a sparse constraint mechanism based on information gain rate for training; Based on the low-dimensional representation, an initial population is generated using chaotic mapping, and individuals in the population correspond to different combinations of operation strategies; Construct a multi-objective fitness function that comprehensively considers the total revenue of the energy station, user satisfaction and energy utilization rate, and evaluate the fitness of the initial population; Use chaos search to generate disturbance vectors in the neighborhood of the initial population individuals, and update the population by comparing the fitness of the new individuals with the original individuals to achieve global exploration; Based on the population after chaotic search, a multi-objective game mechanism is introduced. Individuals generate new individuals in the neighborhood, play games based on different operating target benefits, adjust their own strategy parameters, and promote population evolution. After multiple iterations of chaotic search and multi-objective game, when the termination condition is met, the strategy corresponding to the individual with the highest fitness in the population is selected to generate the optimal operation strategy including charging pile pricing, energy storage battery charging and discharging, and user discount push.

[0053] In this embodiment, an intelligent decision-making system from high-dimensional feature vectors to optimal operation strategies is constructed to solve the problem of multi-source heterogeneous data processing and complex strategy optimization in the operation of photovoltaic storage and charging integrated energy stations. Among them, for the operation of photovoltaic storage and charging integrated energy stations, starting from high-dimensional feature vectors, the optimal operation strategy is generated through dimensionality reduction and global search. Innovation is reflected in multiple key links. When reducing the dimension, the high-dimensional feature vectors are pre-processed based on the data distribution characteristics, and the information topology map is constructed to clearly present the data association structure. The unique probability activation function can accurately capture the probability distribution of data features. The sparse constraint mechanism based on the information gain rate removes redundant information, making the generated low-dimensional representation more representative and effectively reducing the computational complexity. In terms of generating the optimal operation strategy, the initial population is generated by using chaotic mapping, breaking the limitations of traditional random initialization and making the population distribution more global and diverse. A multi-objective fitness function is constructed by comprehensively considering the total revenue of the energy station, user satisfaction and energy utilization rate to comprehensively measure the pros and cons of the operation strategy. Chaotic search generates disturbance vectors in the neighborhood to update the population, achieve global exploration, and avoid falling into local optimality; the multi-objective game mechanism simulates the competition and cooperation between different operating objectives, and promotes the continuous evolution and optimization of the population. In summary, this embodiment uses innovative dimensionality reduction and global search methods to efficiently extract key information from complex high-dimensional data. The generated optimal operation strategy can comprehensively balance the various interests of the energy station, improve operational efficiency and economic benefits, and enhance the adaptability of the energy station in different scenarios.

[0054] Specifically, compared with the prior art, this embodiment has the following technical improvements: The following technical improvements were made during the dimensionality reduction phase: Preprocessing based on data distribution characteristics: Traditional data preprocessing methods often use a unified standardization or normalization method without fully considering the actual distribution of the data. This embodiment deeply analyzes the distribution characteristics of the data, such as normal distribution, skewed distribution, etc., and adopts specific preprocessing methods for different distribution types. For example, for skewed distributed data, it is converted into a form closer to normal distribution through methods such as logarithmic transformation or Box-Cox transformation, so that the data has better stability and comparability in subsequent processing, avoiding feature extraction bias caused by uneven data distribution.

[0055] Constructing an information topology map: Previous data processing rarely analyzes the relationship between data from the perspective of topological structure. This embodiment innovatively constructs an information topology map, treating each feature in a high-dimensional feature vector as a node in the graph, and the strength of the association between features is represented by the weight of the edge. This topological structure can intuitively display the complex dependencies between data and provide richer information for subsequent encoding and feature extraction. For example, by analyzing the information topology map, it can be found that there is a potential correlation between the power generation of photovoltaic panels and weather factors (such as light intensity, temperature) and the frequency of use of charging piles, which helps to capture key information in the data more accurately.

[0056] Unique probabilistic activation function: Traditional activation functions (such as Sigmoid, ReLU, etc.) are deterministic and cannot effectively handle uncertainty in data. This embodiment introduces a unique probabilistic activation function, which is based on a probabilistic model and can dynamically adjust the activation state of neurons according to the probability distribution of input data. For example, when processing photovoltaic panel power generation data, taking into account the uncertainty of weather, the probabilistic activation function can determine the degree of neuron activation based on the probability of power generation under different weather conditions, thereby more accurately reflecting the characteristics and laws of the data and improving the accuracy and robustness of the encoding.

[0057] Sparse constraint mechanism based on information gain rate: The traditional sparse constraint mechanism mainly focuses on the sparsity of data and ignores the information value of features. This embodiment performs sparse constraints based on the information gain rate, which measures the contribution of each feature to the overall data information. During the training process, by constraining features with low information gain rates, redundant information is removed and the most critical features for generating operational strategies are retained. This not only reduces the dimensionality of the data and the amount of calculation, but also improves the quality of low-dimensional representation, making subsequent strategy searches more efficient.

[0058] The following technical improvements were made in the phase of generating the optimal operating strategy: Generating initial population using chaotic mapping: Traditional genetic algorithms or evolutionary algorithms usually generate initial populations by random initialization, which easily leads to uneven population distribution and falls into local optimal solutions. This embodiment utilizes the randomness, ergodicity and sensitivity to initial conditions of chaotic mapping to generate a wider and more evenly distributed initial population. For example, a set of initial solutions are randomly generated in the search space through chaotic systems such as Logistic mapping or Tent mapping. These solutions can cover a wider area, increasing the possibility of finding a global optimal solution.

[0059] Constructing a multi-objective fitness function: Traditional operation strategy optimization often only considers a single goal, such as the economic benefits or energy utilization rate of the energy station, while ignoring other important factors. This embodiment constructs a multi-objective fitness function that comprehensively considers the total revenue of the energy station, user satisfaction and energy utilization rate. By reasonably setting the weights of each goal, the pros and cons of each operation strategy can be comprehensively evaluated. For example, during the peak tourist season, energy stations may pay more attention to user satisfaction to attract more users; while during periods of energy shortage, energy utilization may become a more important goal. The multi-objective fitness function can dynamically adjust the optimization direction of the strategy according to different scenarios and needs.

[0060] Chaotic search: Traditional search algorithms are prone to falling into local optimal solutions during the search process and cannot find the global optimal solution. This embodiment introduces chaotic search, which generates a perturbation vector in the neighborhood of the initial population individuals and slightly perturbs the current solution, thereby jumping out of the local optimal trap and achieving global exploration. For example, when searching for charging pile pricing strategies, chaotic search can make slight adjustments based on the current pricing and try different price combinations to find the optimal pricing that can both increase revenue and meet user needs.

[0061] Multi-objective game mechanism: The operation of energy stations in reality involves multiple stakeholders and multiple operational goals, and there are often conflicts and competitions between these goals. This embodiment introduces a multi-objective game mechanism to simulate the game process between different operational goals. After each individual generates a new individual in the neighborhood, it plays a game based on the benefits of different operational goals and dynamically adjusts its own strategy parameters. For example, when formulating a storage battery charging and discharging strategy, the two goals of maximizing benefits and maximizing energy utilization may conflict. Through the multi-objective game mechanism, individuals can find a balance between the two, promote the continuous evolution of the population, and generate better operational strategies.

[0062] In summary, in the dimensionality reduction stage, the generated low-dimensional representation can accurately retain the key information in the high-dimensional feature vector, while reducing the dimension of the data, reducing the computational complexity, and improving the efficiency of subsequent strategy search. In the stage of generating the optimal operation strategy, it can comprehensively balance the various interests of the energy station. The generated strategy takes into account both the economic benefits of the energy station and user satisfaction and energy utilization, improving the operational efficiency and competitiveness of the energy station, so that it can better adapt to different market environments and user needs.

[0063] Specifically, the working process of this embodiment is as follows: (1) Improved sparse autoencoder dimensionality reduction based on information topology and probabilistic activation: Step 1: Data preprocessing and distribution adaptation: Obtain the high-dimensional feature vector output from step S12 , these vectors integrate the complex dependencies of multi-source heterogeneous data of the integrated energy station of photovoltaic storage and charging. In order to solve the scale differences and distribution characteristics between different features, we first calculate each feature Skewness and Kurtosis Skewness Used to measure the asymmetry of data distribution, kurtosis Reflects the steepness of the data distribution. The mean of each feature is obtained through statistical calculation and standard deviation According to the formula Each feature is standardized. This formula not only unifies the data scale, but also fine-tunes the standardization process according to the data distribution characteristics. For example, for features with large skewness, This item modifies the normalization result to make it more in line with the requirements of subsequent encoding operations, thereby obtaining the preprocessed feature vector .

[0064] Step 2: Construct information topology and coding layer calculation: The preprocessed feature vector The information topology is constructed based on the information topology. The nodes in the graph correspond to each feature, and the weight of the edge is determined by the mutual information between the features. Mutual information is used to measure the degree of dependence between two features. By calculating the mutual information between different feature pairs, the strength of the association between the features can be clarified. Layer encoding layer, neurons Output According to the formula Calculate. Among them, It is Layer from Neuron to The weight of a neuron, It is The number of neurons in the layer, (number of neurons in the input layer); It is Tier The bias of each neuron; is a newly designed probability activation function, defined as , is a random number that follows a standard normal distribution, is a parameter that controls the degree of random perturbation (e.g. ). Representation characteristics and In this way, the encoding layer can dynamically adjust the weight of information transmission according to the correlation strength between features. At the same time, the randomness introduced by the probabilistic activation function helps the model to jump out of the local optimal solution, thus obtaining the first The output of the layer encoding layer .

[0065] Step 3: Strengthen sparse constraints and training: In order to highlight the key information, a constraint based on information gain rate is introduced on the basis of traditional sparse constraints. Tier neurons, calculate their information gain rate The information gain rate is obtained by calculating the contribution of the neuron to the reduction of reconstruction error under different input data. Minimize the objective function: Perform model training. is the number of training samples, It is input samples, is the decoded output sample; is the weight of the sparse penalty term (e.g. ), used to balance reconstruction error and sparsity; is the expected information gain rate value, which is set according to the data complexity and key information extraction target. By continuously adjusting the weight parameters of the encoding layer and the decoding layer, the objective function value is gradually reduced, prompting the neurons to focus on the key information that has the greatest impact on the reconstruction error while ensuring the reconstruction accuracy. After multiple rounds of training, the output of the innermost hidden layer is the low-dimensional representation after dimensionality reduction. .

[0066] (2) Improved hybrid evolutionary algorithm generation strategy based on chaotic search and multi-objective game: Step 1: Initial population generation and strategy mapping: Represented in low dimensional form after dimensionality reduction Generate the initial population based on (For example ). Each individual Corresponding to a combination of operating strategies. For charging pile pricing strategies, individual codes include the proportional factors of price adjustments in different time periods (such as peak, flat, and valley); energy storage battery charging and discharging strategy codes cover charging and discharging decisions under different power thresholds; user discount push strategy codes include discount types (such as discounts, free charging time), discount strength, and push timing. Chaotic mapping is used to generate initial values ​​to ensure the diversity of the initial population. For example, using Logistic chaotic mapping ,in Take 4 (in a chaotic state). Generate a series of chaotic sequences through multiple iterations and map them to the value range of the corresponding strategy parameters. Assuming that the value range of the charging pile peak period price adjustment factor is [0.8, 1.2], transform the chaotic sequence value through linear transformation (in , is the chaotic sequence value) is mapped into this range to obtain the initial population individuals and form the initial population .

[0067] Step 2: Construction and evaluation of multi-objective fitness function: Designing fitness function To evaluate the pros and cons of each individual, and comprehensively consider multiple operational goals of the energy station. Including the total revenue of the energy station , user satisfaction and energy efficiency .

[0068] Total revenue It is obtained by calculating the charging pile income, photovoltaic power generation sales income, and the energy storage battery charge and discharge price difference. For example, the charging pile income is calculated based on the charging price and charging amount in different periods, the photovoltaic power generation sales income is calculated based on the power generation and the grid purchase price, and the energy storage battery charge and discharge price difference is calculated based on the charge and discharge amount and the electricity price difference. User satisfaction Quantitative evaluation based on factors such as user charging waiting time and charging price acceptance. Through questionnaire surveys or user behavior data statistics, user feedback on waiting time and price can be converted into corresponding satisfaction scores. Energy utilization rate It is measured by indicators such as photovoltaic power consumption rate and energy storage battery charging and discharging efficiency. The photovoltaic power consumption rate is the ratio of photovoltaic power generation used in the station and transmitted to the power grid to the total power generation, and the energy storage battery charging and discharging efficiency is the ratio of discharged power to charged power.

[0069] The fitness function is expressed as: , , are the minimum and maximum values ​​of each indicator in the training samples; is the weight coefficient (e.g. ), and adjust according to the actual operation focus of the energy station to balance the importance of different indicators in the optimization process. Each individual in Calculate its fitness value and get the fitness evaluation result.

[0070] Step 3: Chaos Search Phase At the beginning of the algorithm, chaos search is used for global exploration. , a set of chaotic perturbation vectors is generated through chaotic mapping For example, using Tent Chaos Map (in ), generate a series of chaotic values. Map these chaotic values ​​according to the change range of the strategy parameters to obtain the chaotic disturbance vector Assuming that the price adjustment factor in the off-peak period in the charging pile pricing strategy varies within the range of [-0.1, 0.1], the chaotic value is transformed linearly. (in , is a chaotic value) is mapped to this range to obtain the disturbance value corresponding to the strategy parameter, and then form a chaotic disturbance vector .

[0071] Get new individuals ,calculate Fitness .if , then the updated individual is By continuously performing chaotic search and updating on individuals in the population, the algorithm can quickly traverse the solution space in the initial stage, find a better area, and obtain the population after chaotic search and update. .

[0072] Step 4: Multi-objective game optimization stage: After the chaotic search, a multi-objective game mechanism is introduced. The individuals in the population are regarded as game participants, and different operation goals are regarded as different strategies of the game. Each individual in , generating a series of new individuals in its neighborhood The generation of the neighborhood can be done by For example, for the charging pile pricing strategy parameter, around its current value Randomly generate new values ​​within the range of .

[0073] Individuals compete by comparing their gains under different goals. For example, under the total gain goal, and Compare and ; Under the user satisfaction goal, compare and Etc. According to the game results, individuals adjust their own strategy parameters and evolve in a better direction. For example, if individuals Outperforms under total return target , but inferior to , then according to the gap between the two and the weight coefficient, The charging pile pricing strategy and user discount push strategy parameters are adjusted. Assuming the total revenue weight , user satisfaction weight ,if , , then according to the formula Calculate the adjustment amount of charging pile pricing strategy parameters (Assuming that this parameter has a linear relationship with benefits and satisfaction), The relevant parameters are adjusted to obtain the population after multi-objective game optimization. .

[0074] Step 5: Iteration and strategy determination After multiple alternating iterations of chaotic search and multi-objective game, steps 3 and 4 in the process of "improved hybrid evolutionary algorithm generation strategy based on chaotic search and multi-objective game" are repeated. Each iteration can enable the individuals in the population to continuously evolve in a better direction. As the number of iterations increases, the population gradually converges. When certain termination conditions are met (for example, the fitness value of the best individual in the population has not increased significantly after several consecutive iterations), the operation strategy corresponding to the individual with the highest fitness in the population is the generated optimal operation strategy. These strategies cover charging pile pricing strategies, energy storage battery charging and discharging strategies, and user discount push strategies, etc., to achieve efficient operation of integrated energy stations with light storage and charging.

[0075] In order to facilitate the understanding of this embodiment, the following example is given here: Assume that a photovoltaic storage and charging integrated energy station is located in an industrial center. The area has a large flow of people, the charging demand of electric vehicles fluctuates significantly, and the power grid electricity price varies greatly in different time periods. After obtaining the high-dimensional feature vector reflecting the operation of the energy station, preprocessing based on the data distribution characteristics is performed. For example, it is found that the frequency of use of charging piles shows an obvious skewed distribution during the morning and evening peaks on weekdays, and it is converted into data closer to a normal distribution through logarithmic transformation for subsequent processing. When constructing the information topology map, the characteristics such as photovoltaic panel power generation, energy storage battery power, charging pile use frequency, and power grid electricity price are used as nodes. Through analysis, it is found that the photovoltaic panel power generation has a strong correlation with the light intensity and time, and the charging pile use frequency is also related to the power grid electricity price and time. Strong edges are established between the corresponding nodes in the information topology map. A unique probability activation function encoding is adopted, taking into account the impact of weather changes on photovoltaic panel power generation, and dynamically activating neurons according to the probability of power generation under different weather conditions, which more accurately reflects the data characteristics. A sparse constraint mechanism based on information gain rate is introduced to remove some features that have little impact on the operation strategy, such as occasional equipment failure information, and generate a low-dimensional representation. Based on the low-dimensional representation, the initial population is generated using chaotic mapping, and each individual represents a combination of operation strategies, including charging pile pricing strategy, energy storage battery charging and discharging strategy, and user preferential push strategy. A multi-objective fitness function is constructed, and the weights for the total revenue of the energy station, user satisfaction, and energy utilization rate are set to 0.5, 0.3, and 0.2, respectively. In the chaotic search stage, the neighborhood of each individual is disturbed, such as fine-tuning the charging pile pricing or the charging and discharging time of the energy storage battery. For example, during the low-price period of the power grid, the charging amount of the energy storage battery is appropriately increased; during the peak period, the pricing of the charging pile is increased and some preferential activities are launched to balance the revenue and user satisfaction. The population is updated by comparing the fitness of the new individual with the original individual. In the multi-objective game mechanism, the strategies represented by different individuals compete to achieve their respective goals. For example, the strategy represented by one individual focuses more on the revenue of the energy station, while the strategy represented by another individual focuses more on user satisfaction. During the game, they will dynamically adjust their own strategy parameters according to the benefits of different operating objectives. After multiple iterations of chaotic search and multi-objective game, when the termination condition is met, the strategy corresponding to the individual with the highest fitness in the population is selected. The optimal operating strategy generated in the end may be to purchase electricity from the power grid at a lower price to charge the energy storage battery during the low-price period of the power grid; during peak hours, increase the pricing of charging piles, but at the same time launch preferential activities such as recharge rebates to attract users; energy storage batteries cooperate with photovoltaic panels to power charging piles during peak hours, improve energy utilization, and achieve the best balance between energy station revenue, user satisfaction and energy utilization.

[0076] As one example of the above scheme, the optimal operation strategy is applied in real time to the operation of the photovoltaic storage and charging integrated energy station; wherein, according to the charging pile pricing strategy and through the smart meter, the pricing of the charging piles of the photovoltaic storage and charging integrated energy station is dynamically adjusted, according to the user preferential push strategy, discount information is pushed to the user terminal, and according to the energy storage battery charging and discharging strategy and through the energy storage management system, the charging and discharging strategy of the energy storage battery of the photovoltaic storage and charging integrated energy station is optimized, including the following steps: (1) Smart meters dynamically adjust charging pile pricing: Real-time data collection: Smart meters continuously collect data from multiple sources, including real-time grid electricity prices, which are obtained in real time from grid operators through a dedicated communication interface, and can accurately reflect changes in grid-side electricity costs at different times; real-time monitoring of photovoltaic panel power generation, through the data line connected to the photovoltaic panel control system, to obtain power generation data per second, to understand the real-time supply of photovoltaic energy; and real-time power of energy storage batteries, through communication with the energy storage management system, to clearly understand the amount of energy that can be allocated by energy storage; at the same time, record the real-time power load of the charging pile, obtain real-time power data from the power monitoring module inside the charging pile, and know the current charging demand intensity.

[0077] Pricing strategy matching: The collected data is matched and analyzed with the charging pile pricing strategy generated by the improved hybrid evolutionary algorithm. For example, if the pricing strategy stipulates that the charging pile electricity price should be increased when the grid electricity price is high and the photovoltaic power generation is insufficient, the smart meter determines that the current grid electricity price is higher than the set threshold and the photovoltaic power can only meet 30% of the electricity demand in the station, then the price increase mechanism is triggered. If the strategy is to reduce the electricity price when the electricity consumption is low and the energy storage power is sufficient, when the smart meter detects that the night grid electricity price is at a low period and the energy storage battery power reaches more than 80%, the price reduction operation is executed.

[0078] Price update: The smart meter calculates the adjusted charging pile price based on the analysis results and sends the new price information to the charging pile control unit through a standard communication protocol (such as ModbusTCP). After receiving the command, the control unit quickly updates the price information on the charging pile operation interface and display screen to ensure that users can see the latest pricing when connecting to the charging device.

[0079] (2) Push discount information to user terminals: User portrait construction: The energy station operation platform integrates the user's registration information, historical charging records and other data on the platform. Through data mining technology, analyze the user's charging time pattern, such as some users are used to charging before going to work on weekday mornings, while others often charge on weekend evenings; count the charging frequency to distinguish between high-frequency users (such as charging more than 3 times a week) and low-frequency users; calculate the charging power to understand the scale of electricity demand of different users. Combine this information to build a detailed portrait for each user.

[0080] Discount plan formulation: Generate personalized discount information based on user portraits and established user discount push strategies. For users who charge frequently and in large quantities, coupons may be provided to reduce certain amounts of money, such as "50 yuan off for purchases over 200 yuan"; for users who often charge during low electricity consumption periods, discount ratios may be given, such as "80% discount for charging during low electricity consumption periods"; for newly registered users, activities such as "10 yuan off for first charge for new users" may be launched to reduce a certain amount of money for the first charge.

[0081] Information push implementation: The generated discount information will be accurately pushed to the corresponding users through the push service of the official mobile application of the energy station. For users who have not installed the application, the discount code and instructions for use will be sent via SMS. After the user receives the information, he can enter the discount code in the application or on the charging pile operation interface to enjoy the corresponding discount the next time he charges.

[0082] (3) Energy storage management system optimizes the charging and discharging strategy of energy storage batteries: Battery status monitoring: The energy storage management system monitors multiple key parameters of the energy storage battery in real time. Through built-in sensors and communication modules, the remaining capacity (SOC) of the battery is obtained, accurate to two decimal places, reflecting the energy storage level of the battery in real time; the battery voltage is monitored to ensure that it is within the normal working range to prevent overvoltage or undervoltage from damaging the battery; the battery temperature is measured, because the temperature has a significant impact on the performance and life of the battery, and the average temperature is obtained through temperature sensors distributed in different positions of the battery pack; the charge and discharge current is monitored, and the current size is controlled to ensure the safe and stable operation of the battery.

[0083] Execution of charging and discharging strategies: Based on the charging and discharging strategies generated by the improved algorithm, the energy storage management system formulates a specific execution plan. When the photovoltaic power generation is higher than the power demand in the station and the battery power is less than 80%, the system controls the battery to charge with an appropriate charging current, such as setting it to 0.2C of the rated capacity of the battery (C is the battery capacity), which can not only ensure the charging efficiency, but also avoid damage to the battery due to excessive current. When the power price of the grid is high and the photovoltaic power generation is insufficient, the energy storage battery is used to power the charging pile first, and the discharge current is adjusted according to the real-time power demand. For example, during the peak power consumption period, if the total power demand of the charging pile is 50kW, the energy storage system will meet the power demand with a stable current output according to the current state of the battery.

[0084] Dynamic adjustment of strategies: The energy storage management system continuously evaluates the execution effect of the charging and discharging strategies. If the battery temperature rises too quickly during the charging process and exceeds the set safety temperature threshold (such as 40°C), the system automatically reduces the charging current to prevent the battery from overheating; if the power demand of the energy station suddenly increases and exceeds the current power supply of the energy storage battery, the system quickly adjusts the discharge strategy and tries to supplement part of the power from the grid while ensuring battery safety to ensure stable power supply of the charging pile.

[0085] See also Figure 2 , is a schematic diagram of a multi-strategy operation device for a photovoltaic storage and charging station based on deep learning provided by an embodiment of the present invention. The multi-strategy operation device for a photovoltaic storage and charging station based on deep learning comprises: The data acquisition module 10 is used to acquire the multi-source heterogeneous operation data of the photovoltaic storage and charging integrated energy station, wherein the operation data includes the status data of the photovoltaic panels, the energy storage batteries and the usage status data of the charging piles; A data fusion module 11, configured to perform data fusion on the operation data based on an improved spatiotemporal heterogeneous graph fusion model and using an improved gated graph convolutional network to obtain fused data; The feature extraction module 12 is used to perform feature extraction operations on the fused data to obtain a high-dimensional feature vector that comprehensively reflects the complex dependency relationship between multi-source heterogeneous operation data; A strategy generation module 13 is used to reduce the dimension and extract key information of the high-dimensional feature vector by an improved sparse autoencoder to generate a low-dimensional representation, and to perform a global search on the low-dimensional representation by an improved hybrid evolutionary algorithm to generate an optimal operation strategy; the optimal operation strategy includes a charging pile pricing strategy, an energy storage battery charging and discharging strategy, and a user discount push strategy; The strategy application module 14 is used to apply the optimal operation strategy to the operation of the photovoltaic, storage and charging integrated energy station in real time; wherein, the pricing of the charging piles of the photovoltaic, storage and charging integrated energy station is dynamically adjusted according to the charging pile pricing strategy and through the smart meter, discount information is pushed to the user terminal according to the user preferential push strategy, and the charging and discharging strategy of the energy storage battery of the photovoltaic, storage and charging integrated energy station is optimized according to the energy storage battery charging and discharging strategy and through the energy storage management system.

[0086] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: The embodiments of the present invention use innovative data fusion, feature extraction and strategy generation technologies to deeply mine the complex dependencies between multi-source heterogeneous operation data, accurately and in real time formulate and apply the optimal operation strategy, and significantly improve the energy utilization efficiency and intelligent operation level of the integrated photovoltaic storage and charging energy station.

[0087] As one example of the above solution, the data fusion module is specifically used for: The operation data is normalized using a normalization algorithm that combines logarithmic transformation and linear scaling, thereby mapping the operation data to the adaptation interval to obtain the preprocessed operation data; According to the preprocessed operation data, the photovoltaic panels, energy storage batteries and charging piles of the photovoltaic storage and charging integrated energy station are abstracted into nodes of a spatiotemporal heterogeneous graph and their attributes are expanded. At the same time, the time edges and space edges of the spatiotemporal heterogeneous graph are defined, and the weights of each edge of the spatiotemporal heterogeneous graph are calculated to reflect the spatiotemporal correlation between the nodes, and the spatiotemporal heterogeneous graph is obtained by composing the graph; By using an improved gated graph convolutional network, a multi-layer network is operated layer by layer on the constructed spatiotemporal heterogeneous graph to aggregate the local features of the nodes, thereby obtaining fused data.

[0088] As one example of the above solution, the feature extraction module is specifically used for: Reshape the fused data and build it into a data cube structure according to different data sources and time series. At the same time, dynamically divide the windows according to the data fluctuation conditions to generate adaptive dynamic window data. Based on the obtained dynamic window data, the comprehensive importance of each feature is evaluated in each window from the information entropy and feature correlation level, and the important features are screened out to form a screened feature subset; For the filtered feature subset, a deep residual attention network is built, and the residual connection is used to solve the gradient problem. The attention mechanism is used to strengthen the expression of key features. After processing, global average pooling is performed to obtain a high-dimensional feature vector that comprehensively reflects the complex dependencies between multi-source heterogeneous operation data.

[0089] As one example of the above solution, the strategy generation module is specifically used for: The high-dimensional feature vector is first preprocessed based on data distribution characteristics, and a low-dimensional representation after dimensionality reduction is generated by constructing an information topology map and encoding it using a unique probability activation function, while introducing a sparse constraint mechanism based on information gain rate for training; Based on the low-dimensional representation, an initial population is generated using chaotic mapping, and individuals in the population correspond to different combinations of operation strategies; Construct a multi-objective fitness function that comprehensively considers the total revenue of the energy station, user satisfaction and energy utilization rate, and evaluate the fitness of the initial population; Use chaos search to generate disturbance vectors in the neighborhood of the initial population individuals, and update the population by comparing the fitness of the new individuals with the original individuals to achieve global exploration; Based on the population after chaotic search, a multi-objective game mechanism is introduced. Individuals generate new individuals in the neighborhood, play games based on different operating target benefits, adjust their own strategy parameters, and promote population evolution. After multiple iterations of chaotic search and multi-objective game, when the termination condition is met, the strategy corresponding to the individual with the highest fitness in the population is selected to generate the optimal operation strategy including charging pile pricing, energy storage battery charging and discharging, and user discount push.

[0090] See also Figure 3 , is a schematic diagram of a multi-strategy operation device for a photovoltaic charging station based on deep learning provided by an embodiment of the present invention. The multi-strategy operation device for a photovoltaic charging station based on deep learning of this embodiment includes: a processor 100, a memory 101, and a computer program stored in the memory 101 and executable on the processor 100, such as a multi-strategy operation program for a photovoltaic charging station based on deep learning. When the processor 100 executes the computer program, the steps in each of the above-mentioned embodiments of the multi-strategy operation method for a photovoltaic charging station based on deep learning are implemented. Alternatively, when the processor 100 executes the computer program, the functions of each module / unit in the above-mentioned device embodiments are implemented.

[0091] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the multi-strategy operation device for photovoltaic storage and charging stations based on deep learning.

[0092] The multi-strategy operation equipment for the photovoltaic storage and charging station based on deep learning can be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The multi-strategy operation equipment for the photovoltaic storage and charging station based on deep learning may include, but is not limited to, a processor and a memory. Those skilled in the art will understand that the schematic diagram is only an example of a multi-strategy operation equipment for the photovoltaic storage and charging station based on deep learning, and does not constitute a limitation on the multi-strategy operation equipment for the photovoltaic storage and charging station based on deep learning. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the multi-strategy operation equipment for the photovoltaic storage and charging station based on deep learning may also include input and output devices, network access equipment, buses, etc.

[0093] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the multi-strategy operation device of the optical storage and charging station based on deep learning, and uses various interfaces and lines to connect the various parts of the multi-strategy operation device of the optical storage and charging station based on deep learning.

[0094] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the optical storage and charging station multi-strategy operation device based on deep learning by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0095] Wherein, if the module / unit integrated with the multi-strategy operation equipment of the optical storage and charging station based on deep learning is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media does not include electrical carrier signals and telecommunication signals.

[0096] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0097] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A multi-strategy operation method for a photovoltaic storage and charging station based on deep learning, characterized in that: include: Acquire the multi-source heterogeneous operation data of the photovoltaic storage and charging integrated energy station, which includes the status data of photovoltaic panels, energy storage batteries and the usage status data of charging piles; Based on an improved spatiotemporal heterogeneous graph fusion model and using an improved gated graph convolutional network, data fusion is performed on the operation data to obtain fused data; Perform feature extraction on the fused data to obtain a high-dimensional feature vector that comprehensively reflects the complex dependencies between multi-source heterogeneous operational data; The high-dimensional feature vector is reduced in dimension and key information is extracted by an improved sparse autoencoder to generate a low-dimensional representation, and the low-dimensional representation is globally searched by an improved hybrid evolutionary algorithm to generate an optimal operation strategy; the optimal operation strategy includes a charging pile pricing strategy, an energy storage battery charging and discharging strategy, and a user discount push strategy; The optimal operation strategy is applied in real time to the operation of the photovoltaic, storage and charging integrated energy station; wherein, the pricing of the charging piles of the photovoltaic, storage and charging integrated energy station is dynamically adjusted according to the charging pile pricing strategy and through smart meters, discount information is pushed to user terminals according to the user preferential push strategy, and the charging and discharging strategy of the energy storage battery of the photovoltaic, storage and charging integrated energy station is optimized according to the energy storage battery charging and discharging strategy and through an energy storage management system.

2. The multi-strategy operation method of a photovoltaic storage and charging station based on deep learning as claimed in claim 1, characterized in that: The improved spatiotemporal heterogeneous graph fusion model is used to fuse the operation data using an improved gated graph convolutional network to obtain fused data, including: The operation data is normalized using a normalization algorithm that combines logarithmic transformation and linear scaling, thereby mapping the operation data to the adaptation interval to obtain the preprocessed operation data; According to the preprocessed operation data, the photovoltaic panels, energy storage batteries and charging piles of the photovoltaic storage and charging integrated energy station are abstracted into nodes of a spatiotemporal heterogeneous graph and their attributes are expanded. At the same time, the time edges and space edges of the spatiotemporal heterogeneous graph are defined, and the weights of each edge of the spatiotemporal heterogeneous graph are calculated to reflect the spatiotemporal correlation between the nodes, and the spatiotemporal heterogeneous graph is obtained by composing the graph; By using an improved gated graph convolutional network, a multi-layer network is operated layer by layer on the constructed spatiotemporal heterogeneous graph to aggregate the local features of the nodes, thereby obtaining fused data.

3. The multi-strategy operation method of a photovoltaic storage and charging station based on deep learning as claimed in claim 1, characterized in that: The feature extraction operation is performed on the fused data to obtain a high-dimensional feature vector that comprehensively reflects the complex dependency relationship between multi-source heterogeneous operation data, including: Reshape the fused data and build it into a data cube structure according to different data sources and time series. At the same time, dynamically divide the windows according to the data fluctuation conditions to generate adaptive dynamic window data. Based on the obtained dynamic window data, the comprehensive importance of each feature is evaluated in each window from the information entropy and feature correlation level, and the important features are screened out to form a screened feature subset; For the filtered feature subset, a deep residual attention network is built, and the residual connection is used to solve the gradient problem. The attention mechanism is used to strengthen the expression of key features. After processing, global average pooling is performed to obtain a high-dimensional feature vector that comprehensively reflects the complex dependencies between multi-source heterogeneous operation data.

4. The multi-strategy operation method of a photovoltaic storage and charging station based on deep learning as claimed in claim 1, characterized in that: The improved sparse autoencoder is used to reduce the dimension of the high-dimensional feature vector and extract key information to generate a low-dimensional representation, and the low-dimensional representation is globally searched by an improved hybrid evolutionary algorithm to generate an optimal operation strategy; the optimal operation strategy includes a charging pile pricing strategy, an energy storage battery charging and discharging strategy, and a user discount push strategy, including: The high-dimensional feature vector is first preprocessed based on data distribution characteristics, and a low-dimensional representation after dimensionality reduction is generated by constructing an information topology map and encoding it using a unique probability activation function, while introducing a sparse constraint mechanism based on information gain rate for training; Based on the low-dimensional representation, an initial population is generated using chaotic mapping, and individuals in the population correspond to different combinations of operation strategies; Construct a multi-objective fitness function that comprehensively considers the total revenue of the energy station, user satisfaction and energy utilization rate, and evaluate the fitness of the initial population; Use chaos search to generate disturbance vectors in the neighborhood of the initial population individuals, and update the population by comparing the fitness of the new individuals with the original individuals to achieve global exploration; Based on the population after chaotic search, a multi-objective game mechanism is introduced. Individuals generate new individuals in the neighborhood, play games based on different operating target benefits, adjust their own strategy parameters, and promote population evolution. After multiple iterations of chaotic search and multi-objective game, when the termination condition is met, the strategy corresponding to the individual with the highest fitness in the population is selected to generate the optimal operation strategy including charging pile pricing, energy storage battery charging and discharging, and user discount push.

5. A multi-strategy operation device for a photovoltaic storage and charging station based on deep learning, characterized in that: include: The data acquisition module is used to obtain the multi-source heterogeneous operation data of the photovoltaic storage and charging integrated energy station, and the operation data includes the status data of photovoltaic panels, energy storage batteries and the usage status data of charging piles; A data fusion module, used for fusing the operation data based on an improved spatiotemporal heterogeneous graph fusion model and using an improved gated graph convolutional network to obtain fused data; The feature extraction module is used to perform feature extraction operations on the fused data to obtain a high-dimensional feature vector that comprehensively reflects the complex dependencies between multi-source heterogeneous operational data; A strategy generation module, which is used to reduce the dimension and extract key information of the high-dimensional feature vector by an improved sparse autoencoder to generate a low-dimensional representation, and to perform a global search on the low-dimensional representation by an improved hybrid evolutionary algorithm to generate an optimal operation strategy; the optimal operation strategy includes a charging pile pricing strategy, an energy storage battery charging and discharging strategy, and a user discount push strategy; A strategy application module is used to apply the optimal operation strategy to the operation of the photovoltaic, storage and charging integrated energy station in real time; wherein, the pricing of the charging piles of the photovoltaic, storage and charging integrated energy station is dynamically adjusted according to the charging pile pricing strategy and through smart meters, discount information is pushed to user terminals according to the user preferential push strategy, and the charging and discharging strategy of the energy storage battery of the photovoltaic, storage and charging integrated energy station is optimized according to the energy storage battery charging and discharging strategy and through an energy storage management system.

6. The multi-strategy operation device for a photovoltaic storage and charging station based on deep learning as claimed in claim 5, characterized in that: The data fusion module is specifically used for: The operation data is normalized using a normalization algorithm that combines logarithmic transformation and linear scaling, thereby mapping the operation data to the adaptation interval to obtain the preprocessed operation data; According to the preprocessed operation data, the photovoltaic panels, energy storage batteries and charging piles of the photovoltaic storage and charging integrated energy station are abstracted into nodes of a spatiotemporal heterogeneous graph and their attributes are expanded. At the same time, the time edges and space edges of the spatiotemporal heterogeneous graph are defined, and the weights of each edge of the spatiotemporal heterogeneous graph are calculated to reflect the spatiotemporal correlation between the nodes, and the spatiotemporal heterogeneous graph is obtained by composing the graph; By using an improved gated graph convolutional network, a multi-layer network is operated layer by layer on the constructed spatiotemporal heterogeneous graph to aggregate the local features of the nodes, thereby obtaining fused data.

7. The multi-strategy operation device for a photovoltaic storage and charging station based on deep learning as claimed in claim 5, characterized in that: The feature extraction module is specifically used for: Reshape the fused data and build it into a data cube structure according to different data sources and time series. At the same time, dynamically divide the windows according to the data fluctuation conditions to generate adaptive dynamic window data. Based on the obtained dynamic window data, the comprehensive importance of each feature is evaluated in each window from the information entropy and feature correlation level, and the important features are screened out to form a screened feature subset; For the filtered feature subset, a deep residual attention network is built, and the residual connection is used to solve the gradient problem. The attention mechanism is used to strengthen the expression of key features. After processing, global average pooling is performed to obtain a high-dimensional feature vector that comprehensively reflects the complex dependencies between multi-source heterogeneous operation data.

8. The multi-strategy operation device for a photovoltaic storage and charging station based on deep learning as claimed in claim 5, characterized in that: The strategy generation module is specifically used for: The high-dimensional feature vector is first preprocessed based on data distribution characteristics, and a low-dimensional representation after dimensionality reduction is generated by constructing an information topology map and encoding it using a unique probability activation function, while introducing a sparse constraint mechanism based on information gain rate for training; Based on the low-dimensional representation, an initial population is generated using chaotic mapping, and individuals in the population correspond to different combinations of operation strategies; Construct a multi-objective fitness function that comprehensively considers the total revenue of the energy station, user satisfaction and energy utilization rate, and evaluate the fitness of the initial population; Use chaos search to generate disturbance vectors in the neighborhood of the initial population individuals, and update the population by comparing the fitness of the new individuals with the original individuals to achieve global exploration; Based on the population after chaotic search, a multi-objective game mechanism is introduced. Individuals generate new individuals in the neighborhood, play games based on different operating target benefits, adjust their own strategy parameters, and promote population evolution. After multiple iterations of chaotic search and multi-objective game, when the termination condition is met, the strategy corresponding to the individual with the highest fitness in the population is selected to generate the optimal operation strategy including charging pile pricing, energy storage battery charging and discharging, and user discount push.

9. A multi-strategy operation device for a photovoltaic storage and charging station based on deep learning, characterized in that: It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the multi-strategy operation method of a photovoltaic storage and charging station based on deep learning as described in any one of claims 1 to 4.

Images