Scheduling method, device and equipment for integration of transmission, use and storage, and storage medium

Abstract

The invention provides a dispatching method, device and equipment for integration of power generation, consumption and storage, and a storage medium, and the method comprises the steps: dividing a target region into a plurality of target sub-regions based on the historical actual weather data of each meteorological station in the target region; performing data prediction processing on each target sub-region based on the historical actual weather data of each target sub-region by using a first hybrid neural network model to obtain predicted photovoltaic power generation capacity; performing data prediction processing on each target sub-region based on the historical energy price of each target sub-region by using a second hybrid neural network model to obtain a predicted power consumption demand; based on the predicted photovoltaic power generation amount and the predicted power consumption demand, determining and based on the predicted storage power of each target sub-region, and setting a power scheduling strategy among the plurality of target sub-regions; compared with the prior art, the technical scheme of the invention can realize optimal distribution of electric power resources through the photovoltaic power generation amount and the power consumption demand amount in the target area.

Description

Technical Field

[0001] The present application relates to the technical field of energy management, and in particular to a scheduling method, device, equipment and storage medium for integrated generation, use and storage. Background Art

[0002] In existing technologies, photovoltaic power generation, as a type of renewable energy, has a random and volatile output, which poses a significant challenge to the stable operation and power balance of the power system. As the proportion of new energy grid-connected continues to increase, the demand for regulation capabilities of the power system is increasing. However, existing technologies are insufficient in utilizing energy storage systems and other flexible regulation resources, which limits the regulation capabilities of the power grid in response to supply and demand fluctuations.

[0003] In addition, traditional optimization scheduling methods often make decisions based solely on current electricity market prices, failing to fully consider forecast information on photovoltaic power generation output and user-side electricity demand, thereby failing to fully tap the regulatory potential and economic benefits of energy storage systems in the power grid, and failing to achieve effective allocation and optimal utilization of power resources. Summary of the invention

[0004] The technical problem to be solved by the present application is to provide a dispatching method, device, equipment and storage medium that integrates power generation, use and storage, so as to achieve optimal allocation of power resources through photovoltaic power generation and power demand in the target area.

[0005] In order to solve the above technical problems, the present application provides a dispatching method integrating generation, use and storage, including:

[0006] Acquire historical actual weather data corresponding to each meteorological station in the target area, and divide the target area based on the historical actual weather data to obtain multiple target sub-areas;

[0007] Acquire historical actual weather data corresponding to each target sub-region, and use a pre-trained first hybrid neural network model to perform data prediction processing on each target sub-region based on the historical actual weather data to obtain predicted photovoltaic power generation corresponding to each target sub-region;

[0008] Obtaining the historical energy price corresponding to each target sub-region, using the pre-trained second hybrid neural network model to perform data prediction processing on each target sub-region based on the historical energy price, to obtain the predicted electricity demand corresponding to each target sub-region;

[0009] Based on the predicted photovoltaic power generation and the predicted power demand, the predicted storage power of each target sub-area is determined, and based on the predicted storage power, a power dispatch strategy between the multiple target sub-areas is set.

[0010] In a possible implementation, the acquiring of historical actual weather data corresponding to each meteorological station in the target area, and dividing the target area based on the historical actual weather data to obtain a plurality of target sub-areas, specifically includes:

[0011] Obtaining historical actual weather data corresponding to each weather station in the target area, and calculating the weather index corresponding to the historical actual weather data in each weather station based on a preset weather index calculation formula;

[0012] The target area is divided into multiple target sub-areas based on the weather index by using the rotational empirical orthogonal decomposition method.

[0013] In a possible implementation, setting a power dispatching strategy among the multiple target sub-areas based on the predicted stored power specifically includes:

[0014] respectively performing correction processing on the predicted storage power corresponding to each target sub-area to obtain a corrected storage power, and classifying the multiple target sub-areas into supply sub-areas and demand sub-areas based on the corrected storage power;

[0015] When it is detected that the supply sub-area and the demand sub-area exist at the same time, and the calculated total supply power corresponding to the supply sub-area is greater than the calculated total demand power corresponding to the demand sub-area, respectively calculating the distance between each demand sub-area and all supply sub-areas, and determining the scheduling strategy between the supply sub-area and the demand sub-area based on the distance;

[0016] When it is detected that the supply sub-area exists and the demand sub-area does not exist, executing a charging strategy for the energy storage system;

[0017] When it is detected that the demand sub-area exists and the supply sub-area does not exist, a discharging strategy is executed on the energy storage system.

[0018] In a possible implementation, the correcting the predicted storage power corresponding to each target sub-area to obtain the corrected storage power specifically includes:

[0019] Calculate the storage power deviation rate according to multiple historical preset storage powers and the historical actual storage power corresponding to each historical preset storage power, and generate a storage power deviation rate data set;

[0020] The average storage power deviation rate corresponding to the storage power deviation rate data set is calculated based on the preset confidence, and the predicted storage power corresponding to each target sub-area is corrected based on the average storage power deviation rate to obtain the corrected storage power.

[0021] In a possible implementation, respectively calculating the distances from each demand sub-area to all supply sub-areas, and determining the scheduling strategy between the supply sub-areas and the demand sub-areas based on the distances, specifically includes:

[0022] Obtaining the power requirements corresponding to each of all the demand sub-areas, and sorting all the demand sub-areas based on the power requirements to obtain a demand sub-area sequence;

[0023] Selecting a single target demand sub-region from the demand sub-region sequence in turn, calculating the distance between the target demand sub-region and all supply sub-regions, and sorting all supply sub-regions in order from near to far based on the distance to obtain a supply sub-region sequence corresponding to the target demand sub-region;

[0024] The supply sub-area sequence is detected in order from left to right. When it is detected that the target supply power corresponding to the target number of target supply sub-areas is greater than the target demand power corresponding to the target demand sub-area, the target number of target supply sub-areas are used as the power scheduling areas of the target demand sub-areas, and a supply sub-area power scheduling strategy is generated. The supply power corresponding to each of the supply sub-areas is updated based on the supply sub-area power scheduling strategy until all the demand sub-areas in the demand sub-area sequence are selected, and the scheduling strategy between the supply sub-areas and the demand sub-areas is determined.

[0025] In a possible implementation, the pre-training process of the first hybrid neural network model specifically includes:

[0026] Constructing an initial first hybrid neural network model, wherein the first hybrid neural network model includes a first convolutional neural network and a first graph neural network, wherein the first convolutional neural network is connected to the first graph neural network;

[0027] Acquire power generation data samples and weather data samples corresponding to the same time, and input the weather data samples into the first hybrid neural network model, so that the first convolutional neural network in the first hybrid neural network model respectively extracts features from the weather data samples to obtain a first feature map matrix;

[0028] Inputting the first feature graph matrix into the first graph neural network based on the first convolutional neural network, so that the first graph neural network performs feature fusion processing on the first feature graph matrix to obtain a first feature fusion graph;

[0029] Inputting the first feature fusion graph into the first convolutional neural network model based on the first graph neural network, so that the first convolutional neural network performs feature extraction on the first feature fusion graph to obtain predicted power generation data;

[0030] The predicted power generation data is compared with the power generation data sample to determine a first hybrid neural network model.

[0031] In a possible implementation, inputting the weather data samples into the first hybrid neural network model so that the first convolutional neural network in the first hybrid neural network model respectively extracts features from the weather data samples to obtain a first feature graph matrix specifically includes:

[0032] The weather data samples include irradiance, cloud cover, humidity and average solar sunshine time;

[0033] Inputting the weather data sample into the first hybrid neural network model, so that the first convolutional neural network in the first hybrid neural network model respectively extracts features of the irradiance, the cloud cover, the humidity and the average solar sunshine time, and respectively obtains feature weights corresponding to the irradiance, the cloud cover, the humidity and the average solar sunshine time;

[0034] Based on the feature weights, an irradiance feature map matrix, a cloud cover feature map matrix, a humidity feature map matrix and an average solar sunshine time feature map matrix are constructed respectively, and the irradiance feature map matrix, the cloud cover feature map matrix, the humidity feature map matrix and the average solar sunshine time feature map matrix are integrated to obtain a first feature map matrix.

[0035] The present invention also provides a dispatching device integrating generation, consumption and storage, comprising: a regional division module, a photovoltaic power generation prediction module, a power demand prediction module and a power dispatching strategy generation module;

[0036] The area division module is used to obtain the historical actual weather data corresponding to each meteorological station in the target area, and divide the target area based on the historical actual weather data to obtain multiple target sub-areas;

[0037] The photovoltaic power generation prediction module is used to obtain the historical actual weather data corresponding to each target sub-region, and use the pre-trained first hybrid neural network model to perform data prediction processing on each target sub-region based on the historical actual weather data to obtain the predicted photovoltaic power generation corresponding to each target sub-region;

[0038] The electricity demand prediction module is used to obtain the historical energy price corresponding to each target sub-area, and use the pre-trained second hybrid neural network model to perform data prediction processing on each target sub-area based on the historical energy price to obtain the predicted electricity demand corresponding to each target sub-area;

[0039] The power dispatch strategy generation module is used to determine the predicted storage power of each target sub-area based on the predicted photovoltaic power generation and the predicted power demand, and to set the power dispatch strategy between the multiple target sub-areas based on the predicted storage power.

[0040] In a possible implementation, the region division module is used to obtain the historical actual weather data corresponding to each meteorological station in the target area, and divide the target area based on the historical actual weather data to obtain multiple target sub-areas, specifically including:

[0041] Obtaining historical actual weather data corresponding to each weather station in the target area, and calculating the weather index corresponding to the historical actual weather data in each weather station based on a preset weather index calculation formula;

[0042] The target area is divided into multiple target sub-areas based on the weather index by using the rotational empirical orthogonal decomposition method.

[0043] In a possible implementation, the power dispatching strategy generating module is used to set the power dispatching strategy between the multiple target sub-areas based on the predicted stored power, specifically including:

[0044] respectively performing correction processing on the predicted storage power corresponding to each target sub-area to obtain a corrected storage power, and classifying the multiple target sub-areas into supply sub-areas and demand sub-areas based on the corrected storage power;

[0045] When it is detected that the supply sub-area and the demand sub-area exist at the same time, and the calculated total supply power corresponding to the supply sub-area is greater than the calculated total demand power corresponding to the demand sub-area, respectively calculating the distance between each demand sub-area and all supply sub-areas, and determining the scheduling strategy between the supply sub-area and the demand sub-area based on the distance;

[0046] When it is detected that the supply sub-area exists and the demand sub-area does not exist, executing a charging strategy for the energy storage system;

[0047] When it is detected that the demand sub-area exists and the supply sub-area does not exist, a discharging strategy is executed on the energy storage system.

[0048] In a possible implementation, the power dispatch strategy generation module is used to perform correction processing on the predicted storage power corresponding to each target sub-area to obtain the corrected storage power, specifically including:

[0049] Calculate the storage power deviation rate according to multiple historical preset storage powers and the historical actual storage power corresponding to each historical preset storage power, and generate a storage power deviation rate data set;

[0050] The average storage power deviation rate corresponding to the storage power deviation rate data set is calculated based on the preset confidence, and the predicted storage power corresponding to each target sub-area is corrected based on the average storage power deviation rate to obtain the corrected storage power.

[0051] In a possible implementation, the power dispatch strategy generation module is used to calculate the distance from each demand sub-area to all supply sub-areas respectively, and determine the dispatch strategy between the supply sub-area and the demand sub-area based on the distance, specifically including:

[0052] Obtaining the power requirements corresponding to each of all the demand sub-areas, and sorting all the demand sub-areas based on the power requirements to obtain a demand sub-area sequence;

[0053] Selecting a single target demand sub-region from the demand sub-region sequence in turn, calculating the distance between the target demand sub-region and all supply sub-regions, and sorting all supply sub-regions in order from near to far based on the distance to obtain a supply sub-region sequence corresponding to the target demand sub-region;

[0054] The supply sub-area sequence is detected in order from left to right. When it is detected that the target supply power corresponding to the target number of target supply sub-areas is greater than the target demand power corresponding to the target demand sub-area, the target number of target supply sub-areas are used as the power scheduling areas of the target demand sub-areas, and a supply sub-area power scheduling strategy is generated. The supply power corresponding to each of the supply sub-areas is updated based on the supply sub-area power scheduling strategy until all the demand sub-areas in the demand sub-area sequence are selected, and the scheduling strategy between the supply sub-areas and the demand sub-areas is determined.

[0055] In a possible implementation, the pre-training process of the first hybrid neural network model in the photovoltaic power generation prediction module specifically includes:

[0056] Constructing an initial first hybrid neural network model, wherein the first hybrid neural network model includes a first convolutional neural network and a first graph neural network, wherein the first convolutional neural network is connected to the first graph neural network;

[0057] Acquire power generation data samples and weather data samples corresponding to the same time, and input the weather data samples into the first hybrid neural network model, so that the first convolutional neural network in the first hybrid neural network model respectively extracts features from the weather data samples to obtain a first feature map matrix;

[0058] Inputting the first feature graph matrix into the first graph neural network based on the first convolutional neural network, so that the first graph neural network performs feature fusion processing on the first feature graph matrix to obtain a first feature fusion graph;

[0059] Inputting the first feature fusion graph into the first convolutional neural network model based on the first graph neural network, so that the first convolutional neural network performs feature extraction on the first feature fusion graph to obtain predicted power generation data;

[0060] The predicted power generation data is compared with the power generation data sample to determine a first hybrid neural network model.

[0061] In a possible implementation, the photovoltaic power generation prediction module is used to input the weather data samples into the first hybrid neural network model so that the first convolutional neural network in the first hybrid neural network model respectively extracts features from the weather data samples to obtain a first feature map matrix, specifically including:

[0062] The weather data samples include irradiance, cloud cover, humidity and average solar sunshine time;

[0063] Inputting the weather data sample into the first hybrid neural network model, so that the first convolutional neural network in the first hybrid neural network model respectively extracts features of the irradiance, the cloud cover, the humidity and the average solar sunshine time, and respectively obtains feature weights corresponding to the irradiance, the cloud cover, the humidity and the average solar sunshine time;

[0064] Based on the feature weights, an irradiance feature map matrix, a cloud cover feature map matrix, a humidity feature map matrix and an average solar sunshine time feature map matrix are constructed respectively, and the irradiance feature map matrix, the cloud cover feature map matrix, the humidity feature map matrix and the average solar sunshine time feature map matrix are integrated to obtain a first feature map matrix.

[0065] The present invention also provides a terminal device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements the integrated dispatching method for generation, use and storage as described in any one of the above items.

[0066] The present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the integrated dispatching method as described in any one of the above.

[0067] The embodiment of the present application provides a dispatching method, device, equipment and storage medium for integrated generation, use and storage, which has the following beneficial effects compared with the prior art:

[0068] By obtaining the historical actual weather data of each meteorological station in the target area and dividing the area, we can more accurately understand the photovoltaic power generation potential and electricity demand of each target sub-area, thereby achieving a more reasonable resource allocation; at the same time, using the pre-trained first hybrid neural network model and the second hybrid neural network model to predict photovoltaic power generation and electricity demand can effectively improve the accuracy of the prediction and reduce the uncertainty caused by weather changes and energy price changes; based on the predicted photovoltaic power generation and electricity demand, the predicted storage power of each target sub-area can be better determined, thereby optimizing the power scheduling strategy, reducing electricity costs, improving economic benefits, and achieving optimal allocation of power resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a flow chart of an embodiment of a dispatching method for integrated generation, use and storage provided by the present application;

[0070] Figure 2 It is a structural schematic diagram of an embodiment of a dispatching device integrating generation, use and storage provided by the present application;

[0071] Figure 3 It is a structural schematic diagram of a terminal device provided by the present invention. DETAILED DESCRIPTION

[0072] The following will be combined with the drawings in this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0073] Example 1, see Figure 1 , Figure 1 : is a flow chart of an embodiment of a dispatching method for integrating generation, use and storage provided by the present application, such as Figure 1 As shown, the method includes steps 101 to 104, which are specifically as follows:

[0074] Step 101: Acquire historical actual weather data corresponding to each meteorological station in a target area, and divide the target area based on the historical actual weather data to obtain multiple target sub-areas.

[0075] In one embodiment, the historical actual weather data corresponding to each meteorological station in the target area is obtained, and based on a preset weather index calculation formula, the weather index corresponding to the historical actual weather data in each meteorological station is calculated; the rotating empirical orthogonal decomposition method is used to divide the target area based on the weather index to obtain multiple target sub-areas.

[0076] Specifically, the historical actual weather data corresponding to each meteorological station is the measured weather data corresponding to a plurality of selected historical moments, and the plurality of historical moments are continuous moments.

[0077] Specifically, the preset weather index calculation formula is as follows:

[0078]

[0079] Where W is the weather index, α is the preset weather weight parameter, T n is the historical actual weather data corresponding to the selected n-th historical moment, and N is the total number of moments corresponding to the selected historical actual weather data.

[0080] In one embodiment, a rotational empirical orthogonal decomposition method is used. When the target area is divided based on the weather index, the weather index corresponding to each meteorological station is subjected to variance calculation to obtain a weather index variance matrix corresponding to each weather index, and the weather index variance matrix is ​​subjected to eigenvalue decomposition to obtain multiple eigenvalues ​​and an eigenvector corresponding to each eigenvalue. Based on the eigenvalues ​​and the eigenvectors corresponding to the eigenvalues, multiple principal components are determined, the multiple principal components are rotated to obtain characteristic modes corresponding to each of the multiple principal components, and the target area is divided based on the characteristic modes to obtain multiple target sub-areas.

[0081] Specifically, when determining multiple principal components based on the eigenvalues ​​and the eigenvectors, the multiple eigenvalues ​​are sorted according to the numerical values ​​corresponding to the multiple eigenvalues, the eigenvectors corresponding to the first n eigenvalues ​​with larger numerical values ​​are selected, and the principal components corresponding to each eigenvector are determined to obtain multiple principal components.

[0082] Step 102: Acquire historical actual weather data corresponding to each target sub-region, and use a pre-trained first hybrid neural network model to perform data prediction processing on each target sub-region based on the historical actual weather data to obtain the predicted photovoltaic power generation corresponding to each target sub-region.

[0083] In one embodiment, the pre-training process of the first hybrid neural network model is specifically carried out by constructing an initial first hybrid neural network model, wherein the first hybrid neural network model includes a first convolutional neural network and a first graph neural network, wherein the first convolutional neural network is connected to the first graph neural network; obtaining power generation data samples and weather data samples corresponding to the same time, and inputting the weather data samples into the first hybrid neural network model, so that the first convolutional neural network in the first hybrid neural network model respectively extracts features from the weather data samples to obtain a first feature graph matrix; based on the first convolutional neural network, the first feature graph matrix is ​​input into the first graph neural network, so that the first graph neural network performs feature fusion processing on the first feature graph matrix to obtain a first feature fusion graph; based on the first graph neural network, the first feature fusion graph is input into the first convolutional neural network model, so that the first convolutional neural network performs feature extraction on the first feature fusion graph to obtain predicted power generation data, and the predicted power generation data is compared with the power generation data samples to determine the first hybrid neural network model.

[0084] In one embodiment, the weather data samples include irradiance, cloud cover, humidity and average solar sunshine time.

[0085] In one embodiment, the first convolutional neural network in the first hybrid neural network model performs feature extraction on the weather data samples respectively. When a first feature graph matrix is ​​obtained, the weather data samples are input into the first hybrid neural network model so that the first convolutional neural network in the first hybrid neural network model performs feature extraction on the irradiance, the cloud cover, the humidity and the average solar sunshine time respectively, and obtains feature weights corresponding to the irradiance, the cloud cover, the humidity and the average solar sunshine time respectively; based on the feature weights, an irradiance feature graph matrix, a cloud cover feature graph matrix, a humidity feature graph matrix and an average solar sunshine time feature graph matrix are respectively constructed, and the irradiance feature graph matrix, the cloud cover feature graph matrix, the humidity feature graph matrix and the average solar sunshine time feature graph matrix are integrated to obtain a first feature graph matrix.

[0086] Preferably, the first convolutional neural network includes a first convolutional layer, and based on the first convolutional layer, feature extraction is performed on the irradiance, the cloud cover, the humidity and the average solar sunshine time respectively.

[0087] Preferably, before inputting the weather data samples into the first hybrid neural network model, the weather data samples are also normalized so as to convert the weather data samples into data in the interval [0,1].

[0088] In one embodiment, the first graph neural network includes a first propagation layer and a second propagation layer, wherein the first propagation layer and the second propagation layer are used to recursively propagate node information.

[0089] In one embodiment, the first graph neural network performs feature fusion processing on the first feature graph matrix. When obtaining the first feature fusion graph, the obtained irradiance feature graph matrix, the cloud cover feature graph matrix, the humidity feature graph matrix and the average solar sunshine time feature graph matrix are respectively input into the first graph neural network, so that the input information of a single node in each feature graph matrix is ​​integrated based on the first propagation layer in the first graph neural network to obtain the node input information, and the output information of a single node in each feature graph matrix is ​​integrated based on the second propagation layer to obtain the node output information, and the node output information and the node input information are integrated based on the attention mechanism to generate the first feature fusion graph.

[0090] In one embodiment, the first convolutional neural network also includes a first entry layer and a second entry layer.

[0091] In one embodiment, the first feature fusion graph generated by the first graph neural network is respectively input into the first convolutional neural network, so that the first entry layer and the second entry layer in the first convolutional neural network perform high-dimensional feature extraction on the first feature fusion graph to obtain the high-dimensional features of the first feature fusion graph, and perform residual connection on all high-dimensional features to obtain predicted power generation data corresponding to each target sub-region.

[0092] In one embodiment, after obtaining the predicted power generation data, the predicted power generation data is also compared with the power generation data sample. If the error between the predicted power generation data and the power generation data sample is less than a preset threshold, it is considered that the trained initial first hybrid neural network model meets the prediction requirements, and the initial first hybrid neural network model is set as the first hybrid neural network model; if the error between the predicted power generation data and the power generation data sample is not less than the preset threshold, the first hybrid neural network model is re-trained based on the power generation data sample and weather data sample corresponding to the same time, until the error between the predicted power generation data and the power generation data sample is less than the preset threshold, and the training of the initial first hybrid neural network model is stopped.

[0093] In one embodiment, after completing the training of the first hybrid neural network model, the acquired historical actual weather data corresponding to each target sub-region is directly output to the first hybrid neural network model, so that the first hybrid neural network model performs data prediction processing on each target sub-region based on the historical actual weather data to obtain the predicted photovoltaic power generation corresponding to each target sub-region.

[0094] Step 103: Obtain the historical energy price corresponding to each target sub-region, and use the pre-trained second hybrid neural network model to perform data prediction processing on each target sub-region based on the historical energy price to obtain the predicted electricity demand corresponding to each target sub-region.

[0095] In one embodiment, the pre-training process of the second hybrid neural network model is specifically carried out by constructing an initial second hybrid neural network model, wherein the second hybrid neural network model includes a second convolutional neural network and a second graph neural network, wherein the second convolutional neural network is connected to the second graph neural network; obtaining electricity consumption data samples and energy price samples corresponding to the same time, and inputting the energy price samples into the second hybrid neural network model, so that the second convolutional neural network in the second hybrid neural network model respectively extracts features from the energy price samples to obtain a second feature graph matrix; based on the second convolutional neural network, the second feature graph matrix is ​​input into the second graph neural network, so that the second graph neural network performs feature fusion processing on the second feature graph matrix to obtain a second feature fusion graph; based on the second graph neural network, the second feature fusion graph is input into the second convolutional neural network model, so that the second convolutional neural network performs feature extraction on the second feature fusion graph to obtain second predicted electricity consumption data; the preset electricity consumption data is compared with the electricity consumption data sample to determine the second hybrid neural network model.

[0096] In one embodiment, the energy price samples include a benchmark price for coal-fired power generation, a market-based transaction electricity price, a peak-valley electricity price, and a green electricity transaction price.

[0097] In one embodiment, the second convolutional neural network in the second hybrid neural network model extracts features from the energy price samples respectively. When a second feature graph matrix is ​​obtained, the energy price samples are input into the second hybrid neural network model so that the second convolutional neural network in the second hybrid neural network model extracts features from the coal-fired power generation benchmark price, the market-based trading electricity price, the peak-valley electricity price and the green electricity trading price respectively, and obtains feature weights corresponding to the coal-fired power generation benchmark price, the market-based trading electricity price, the peak-valley electricity price and the green electricity trading price respectively; based on the feature weights, a coal-fired power generation benchmark price feature graph matrix, a market-based trading electricity price feature graph matrix, a peak-valley electricity price feature graph matrix and a green electricity trading price feature graph matrix are respectively constructed, and the coal-fired power generation benchmark price feature graph matrix, the market-based trading electricity price feature graph matrix, the peak-valley electricity price feature graph matrix and the green electricity trading price feature graph matrix are integrated to obtain a second feature graph matrix.

[0098] Preferably, before inputting the energy price sample into the second hybrid neural network model, the energy price sample is also normalized so as to convert the energy price sample into data in the interval [0,1].

[0099] Preferably, the second convolutional neural network includes a third convolutional layer, and based on the third convolutional layer, features of the coal-fired power generation benchmark price, the market-based trading electricity price, the peak-valley electricity price and the green electricity trading price are respectively extracted.

[0100] In one embodiment, the second graph neural network includes three propagation layers and a fourth propagation layer, wherein the third propagation layer and the fourth propagation layer are used to recursively propagate node information.

[0101] In one embodiment, the second graph neural network performs feature fusion processing on the second feature graph matrix to obtain the second feature fusion graph, and then inputs the obtained coal-fired power generation benchmark price feature graph matrix, the market-based trading electricity price feature graph matrix, the peak-valley electricity price feature graph matrix and the green electricity trading price feature graph matrix into the second graph neural network respectively, so that the input information of a single node in each feature graph matrix is ​​integrated based on the third propagation layer in the second graph neural network to obtain the node input information, and the output information of a single node in each feature graph matrix is ​​integrated based on the fourth propagation layer to obtain the node output information, and the node output information and the node input information are integrated based on the attention mechanism to generate a second feature fusion graph.

[0102] In one embodiment, the second convolutional neural network also includes a third entry layer and a fourth entry layer.

[0103] In one embodiment, the second feature fusion graph generated by the second graph neural network is respectively input into the second convolutional neural network, so that the third entry layer and the fourth entry layer in the second convolutional neural network perform high-dimensional feature extraction on the second feature fusion graph to obtain the high-dimensional features of the second feature fusion graph, and perform residual connection on all high-dimensional features to obtain the predicted electricity consumption data corresponding to each target sub-area.

[0104] In one embodiment, after obtaining the predicted electricity consumption data, the predicted electricity consumption data is also compared with the electricity consumption data sample. If the error between the predicted electricity consumption data and the electricity consumption data sample is less than a preset threshold, it is considered that the trained initial first hybrid neural network model meets the prediction requirements, and the initial second hybrid neural network model is set as the second hybrid neural network model; if the error between the predicted electricity consumption data and the electricity consumption data sample is not less than the preset threshold, the second hybrid neural network model is re-trained based on the electricity consumption data sample and energy price sample corresponding to the same time, until the error between the predicted electricity consumption data and the energy price sample is less than the preset threshold, and the training of the initial second hybrid neural network model is stopped.

[0105] In one embodiment, after completing the training of the second hybrid neural network model, the historical energy prices corresponding to each target sub-region are directly output to the second hybrid neural network model, so that the second hybrid neural network model performs data prediction processing on each target sub-region based on the historical energy prices to obtain the predicted electricity demand corresponding to each target sub-region.

[0106] Preferably, the first hybrid neural network model and the second hybrid neural network model have the same model structure, and the selected first convolutional neural network and the second convolutional neural network are the same network, the first graph neural network and the second graph neural network are also the same network, and the difference between the first hybrid neural network model and the second hybrid neural network model is only that there is a difference in the data used when re-training the model.

[0107] Preferably, the first convolutional neural network and the second convolutional neural network are both CNN models, and the first graph neural network and the second neural network are both GNN models.

[0108] Step 104: Based on the predicted photovoltaic power generation and the predicted power demand, determine the predicted storage power of each target sub-area, and set a power dispatch strategy between the multiple target sub-areas based on the predicted storage power.

[0109] In one embodiment, the predicted photovoltaic power generation and the electricity demand corresponding to each of the target sub-areas are obtained. When the predicted photovoltaic power generation is not less than the electricity demand, the difference between the predicted photovoltaic power generation and the electricity demand is calculated to obtain a first difference, and the first difference is used as the predicted storage power of the target sub-area; when the predicted photovoltaic power generation is less than the electricity demand, the absolute difference between the predicted photovoltaic power generation and the electricity demand is calculated, and the absolute difference is used as the predicted storage power corresponding to the target sub-area.

[0110] In one embodiment, the predicted storage power corresponding to each target sub-region is corrected to obtain a corrected storage power, and based on the corrected storage power, the multiple target sub-regions are classified into supply sub-regions and demand sub-regions.

[0111] Specifically, according to multiple historical preset storage powers and the historical actual storage powers corresponding to each historical preset storage power, the storage power deviation rate is calculated, and a storage power deviation rate data set is generated; the average storage power deviation rate corresponding to the storage power deviation rate data set is calculated based on the preset confidence, and the predicted storage power corresponding to each target sub-area is corrected based on the average storage power deviation rate to obtain the corrected storage power.

[0112] Specifically, when calculating the stored power deviation rate, the multiple historical preset stored powers and the historical actual stored power corresponding to each historical preset stored power are respectively substituted into the preset deviation rate calculation formula to obtain the storage power deviation rates corresponding to each of the multiple historical preset stored powers, wherein the deviation rate calculation formula is as follows:

[0113] ΔQ i =Q Li -Q li / Q Li ×100%;

[0114] Where ΔQ i is the storage power deviation rate corresponding to the i-th historical preset storage power, Q Li is the actual storage capacity of the ith historical data, Q li The storage capacity for the i-th historical preset.

[0115] Specifically, the storage power deviation rates corresponding to the plurality of historical preset storage powers are integrated to obtain a storage power deviation rate data set.

[0116] Specifically, the average storage power deviation rate is determined by setting a preset confidence c; specifically, the preset confidence c is set and based on the preset confidence c, wherein the preset confidence c is the cth percentile value in the storage power deviation rate data set.

[0117] Specifically, when setting and calculating the average storage power deviation rate corresponding to the storage power deviation rate data set based on a preset confidence, the preset confidence is multiplied by the storage power deviation rate in the storage power deviation rate data set to obtain a plurality of storage power deviation rate multiplication values, and the plurality of storage power deviation rate multiplication values ​​are averaged to obtain the average storage power deviation rate corresponding to the storage power deviation rate data set.

[0118] Specifically, when the predicted storage power corresponding to each target sub-area is corrected based on the average storage power deviation rate, the average storage power deviation rate is multiplied by the predicted storage power corresponding to each target sub-area to obtain the corrected storage power corresponding to each target sub-area.

[0119] Specifically, based on the corrected storage power, when the multiple target sub-areas are classified into supply sub-areas and demand sub-areas, it is respectively detected whether the predicted storage power corresponding to each target sub-area is an absolute difference. If so, the target sub-area is classified as a demand sub-area, and the corrected storage power is used as the demand power corresponding to the demand sub-area; if not, the target sub-area is classified as a supply sub-area, and the corrected storage power is used as the supply power corresponding to the supply sub-area.

[0120] In one embodiment, when it is detected that the supply sub-area and the demand sub-area exist at the same time, and the calculated total supply power corresponding to the supply sub-area is greater than the calculated total demand power corresponding to the demand sub-area, the distance from each demand sub-area to all supply sub-areas is calculated respectively, and based on the distance, the scheduling strategy between the supply sub-area and the demand sub-area is determined.

[0121] Specifically, the demand power corresponding to each of all demand sub-areas is obtained, and based on the demand power, all the demand sub-areas are sorted to obtain a demand sub-area sequence; a single target demand sub-area is selected from the demand sub-area sequence in turn, and the distance from the target demand sub-area to all supply sub-areas is calculated. Based on the distance, all the supply sub-areas are sorted in order from near to far to obtain a supply sub-area sequence corresponding to the target demand sub-area; the supply sub-area sequence is detected in order from left to right, and when it is detected that the target supply power corresponding to a target number of target supply sub-areas is greater than the target demand power corresponding to the target demand sub-area, the target number of target supply sub-areas are used as the power scheduling area of ​​the target demand sub-area, and a supply sub-area power scheduling strategy is generated, and the supply power corresponding to each of the supply sub-areas is updated based on the supply sub-area power scheduling strategy, until all demand sub-areas in the demand sub-area sequence are selected, and the scheduling strategy between the supply sub-area and the demand sub-area is determined.

[0122] Preferably, the supply sub-region power dispatching strategy includes a target power supply region, and is aware of the power supplied by the target power supply region to the target demand region.

[0123] In one embodiment, when it is detected that the supply sub-area exists and the demand sub-area does not exist, a charging strategy is executed on the energy storage system.

[0124] Specifically, when it is detected that the supply sub-area exists and the demand sub-area does not exist, the total supply power corresponding to all supply sub-areas is calculated, and the total supply power is used as the input power of the energy storage system to charge the energy storage system.

[0125] In one embodiment, when it is detected that the demand sub-area exists and the supply sub-area does not exist, a discharge strategy is executed on the energy storage system.

[0126] Specifically, when it is detected that the demand sub-area exists and the supply sub-area does not exist, the total power demand corresponding to all demand sub-areas is calculated, and the total power demand is used as the output power of the energy storage system to discharge the energy storage system.

[0127] In one embodiment, when it is detected that the supply sub-area and the demand sub-area exist at the same time, and the calculated total supply power corresponding to the supply sub-area is less than the calculated total demand power corresponding to the demand sub-area, the total demand power difference between the total demand power and the total supply power is calculated, and the total demand power difference is used as the output power of the energy storage system to discharge the energy storage system.

[0128] Example 2, see Figure 2 , Figure 2 : is a structural schematic diagram of an embodiment of a dispatching device for integrating generation, use and storage provided by the present application, such as Figure 2 As shown, the device includes a region division module 201, a photovoltaic power generation prediction module 202, a power demand prediction module 203 and a power dispatch strategy generation module 204, which are specifically as follows:

[0129] The area division module 201 is used to obtain the historical actual weather data corresponding to each meteorological station in the target area, and divide the target area based on the historical actual weather data to obtain multiple target sub-areas.

[0130] The photovoltaic power generation prediction module 202 is used to obtain the historical actual weather data corresponding to each target sub-region, and use the pre-trained first hybrid neural network model to perform data prediction processing on each target sub-region based on the historical actual weather data to obtain the predicted photovoltaic power generation corresponding to each target sub-region.

[0131] The electricity demand prediction module 203 is used to obtain the historical energy price corresponding to each target sub-area, and use the pre-trained second hybrid neural network model to perform data prediction processing on each target sub-area based on the historical energy price to obtain the predicted electricity demand corresponding to each target sub-area.

[0132] The power dispatch strategy generation module 204 is used to determine the predicted storage power of each target sub-area based on the predicted photovoltaic power generation and the predicted power demand, and to set the power dispatch strategy between the multiple target sub-areas based on the predicted storage power.

[0133] In one embodiment, the area division module 201 is used to obtain the historical actual weather data corresponding to each meteorological station in the target area, and divide the target area based on the historical actual weather data to obtain multiple target sub-areas, specifically including: obtaining the historical actual weather data corresponding to each meteorological station in the target area, and calculating the weather index corresponding to the historical actual weather data in each meteorological station based on a preset weather index calculation formula; using the rotating empirical orthogonal decomposition method to divide the target area based on the weather index to obtain multiple target sub-areas.

[0134] In one embodiment, the power dispatch strategy generation module 204 is used to set the power dispatch strategy between the multiple target sub-areas based on the predicted stored power, specifically including: respectively correcting the predicted stored power corresponding to each target sub-area to obtain the corrected stored power, and classifying the multiple target sub-areas into supply sub-areas and demand sub-areas based on the corrected stored power; when it is detected that the supply sub-area and the demand sub-area exist at the same time, and the calculated total supply power corresponding to the supply sub-area is greater than the calculated total demand power corresponding to the demand sub-area, respectively calculating the distance from each demand sub-area to all supply sub-areas, and determining the dispatch strategy between the supply sub-area and the demand sub-area based on the distance; when it is detected that the supply sub-area exists and the demand sub-area does not exist, executing the charging strategy for the energy storage system; when it is detected that the demand sub-area exists and the supply sub-area does not exist, executing the discharging strategy for the energy storage system.

[0135] In one embodiment, the power dispatch strategy generation module 204 is used to perform correction processing on the predicted storage power corresponding to each target sub-area respectively to obtain the corrected storage power, specifically including: calculating the storage power deviation rate according to multiple historical preset storage powers and the historical actual storage power corresponding to each historical preset storage power, and generating a storage power deviation rate data set; setting and calculating the average storage power deviation rate corresponding to the storage power deviation rate data set based on a preset confidence, and performing correction processing on the predicted storage power corresponding to each target sub-area based on the average storage power deviation rate to obtain the corrected storage power.

[0136] In one embodiment, the power dispatch strategy generation module 204 is used to calculate the distance from each demand sub-area to all supply sub-areas respectively, and determine the dispatch strategy between the supply sub-area and the demand sub-area based on the distance, specifically including: obtaining the power demand corresponding to each of the demand sub-areas, and sorting all the demand sub-areas based on the power demand to obtain a demand sub-area sequence; selecting a single target demand sub-area from the demand sub-area sequence in turn, calculating the distance from the target demand sub-area to all supply sub-areas, and sorting all the supply sub-areas in order from near to far based on the distance to obtain the target demand sub-area sequence. a supply sub-region sequence corresponding to a target demand sub-region; detecting the supply sub-region sequence from left to right, and when it is detected that the target supply power corresponding to the target number of target supply sub-regions is greater than the target demand power corresponding to the target demand sub-region, taking the target number of target supply sub-regions as the power dispatching region of the target demand sub-region, and generating a supply sub-region power dispatching strategy, updating the supply power corresponding to each of the supply sub-regions based on the supply sub-region power dispatching strategy, until all demand sub-regions in the demand sub-region sequence are selected, and determining the dispatching strategy between the supply sub-region and the demand sub-region.

[0137] In one embodiment, the pre-training process of the first hybrid neural network model in the photovoltaic power generation prediction module 202 specifically includes: constructing an initial first hybrid neural network model, wherein the first hybrid neural network model includes a first convolutional neural network and a first graph neural network, wherein the first convolutional neural network is connected to the first graph neural network; obtaining power generation data samples and weather data samples corresponding to the same time, and inputting the weather data samples into the first hybrid neural network model, so that the first convolutional neural network in the first hybrid neural network model respectively extracts features from the weather data samples to obtain a first feature graph matrix; based on the first convolutional neural network, inputting the first feature graph matrix into the first graph neural network, so that the first graph neural network performs feature fusion processing on the first feature graph matrix to obtain a first feature fusion graph; based on the first graph neural network, inputting the first feature fusion graph into the first convolutional neural network model, so that the first convolutional neural network performs feature extraction on the first feature fusion graph to obtain predicted power generation data; comparing the predicted power generation data with the power generation data samples to determine the first hybrid neural network model.

[0138] In one embodiment, the photovoltaic power generation prediction module 202 is used to input the weather data samples into the first hybrid neural network model so that the first convolutional neural network in the first hybrid neural network model respectively extracts features from the weather data samples to obtain a first feature map matrix, which specifically includes: the weather data samples include irradiance, cloud cover, humidity and average solar sunshine time; the weather data samples are input into the first hybrid neural network model so that the first convolutional neural network in the first hybrid neural network model respectively extracts features from the irradiance, cloud cover, humidity and average solar sunshine time to obtain feature weights corresponding to the irradiance, cloud cover, humidity and average solar sunshine time; based on the feature weights, an irradiance feature map matrix, a cloud cover feature map matrix, a humidity feature map matrix and an average solar sunshine time feature map matrix are respectively constructed, and the irradiance feature map matrix, the cloud cover feature map matrix, the humidity feature map matrix and the average solar sunshine time feature map matrix are integrated to obtain a first feature map matrix.

[0139] The design device of the integrated dispatching system for generation, use and storage can implement the dispatching method for generation, use and storage in the above method embodiment. The options in the above method embodiment are also applicable to this embodiment and will not be described in detail here.

[0140] Figure 3 FIG. 1 is a schematic diagram of the structure of a terminal device. Figure 3 As shown, the terminal device 3 of this embodiment includes: at least one processor 301 ( Figure 3 Only one is shown in the figure) a processor, a memory 302, and a computer program 303 stored in the memory 302 and executable on at least one processor 301, and when the processor 301 executes the computer program 303, the steps in any of the above method embodiments are implemented.

[0141] The terminal device 3 may be a computing device such as a smart phone, a laptop, a tablet computer, and a desktop computer. The terminal device may include but is not limited to a processor 301 and a memory 302. Those skilled in the art will appreciate that Figure 3 It is only an example of the terminal device 3 and does not constitute a limitation on the terminal device 3. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, it may also include input and output devices, network access devices, etc.

[0142] The processor 301 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 gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0143] In some embodiments, the memory 302 may be an internal storage unit of the terminal device 3, such as a hard disk or memory of the terminal device 3. In other embodiments, the memory 302 may also be an external storage device of the terminal device 3, such as 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), etc. equipped on the terminal device 3. Further, the memory 302 may also include both an internal storage unit and an external storage device of the terminal device 3. The memory 302 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as program codes of a computer program, etc. The memory 302 may also be used to temporarily store data that has been output or is to be output.

[0144] In addition, an embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0145] In several embodiments provided in the present application, it is understood that each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and a module, a program segment or a part of a code includes one or more executable instructions for realizing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes also be executed in the opposite order, depending on the functions involved.

[0146] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a terminal device to execute all or part of the steps of each embodiment method of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code.

[0147] In summary, the present application provides an integrated scheduling method, device, equipment and storage medium for power generation, consumption and storage. Based on the historical actual weather data of each meteorological station in the target area, the target area is divided into multiple target sub-areas; a first hybrid neural network model is used to perform data prediction processing on each target sub-area based on the historical actual weather data of each target sub-area to obtain a predicted photovoltaic power generation; a second hybrid neural network model is used to perform data prediction processing on each target sub-area based on the historical energy price of each target sub-area to obtain a predicted electricity demand; based on the predicted photovoltaic power generation and the predicted electricity demand, the predicted storage power of each target sub-area is determined and a power scheduling strategy between multiple target sub-areas is set; compared with the prior art, the technical solution of the present invention can achieve optimal allocation of power resources through photovoltaic power generation and electricity demand in the target area.

[0148] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and substitutions can be made without departing from the technical principles of the present application. These improvements and substitutions should also be regarded as the scope of protection of the present application.

Claims

1. A dispatching method integrating generation, use and storage, characterized in that: include: Acquire historical actual weather data corresponding to each meteorological station in the target area, and divide the target area based on the historical actual weather data to obtain multiple target sub-areas; Acquire historical actual weather data corresponding to each target sub-region, and use a pre-trained first hybrid neural network model to perform data prediction processing on each target sub-region based on the historical actual weather data to obtain predicted photovoltaic power generation corresponding to each target sub-region; Obtaining the historical energy price corresponding to each target sub-region, using the pre-trained second hybrid neural network model to perform data prediction processing on each target sub-region based on the historical energy price, to obtain the predicted electricity demand corresponding to each target sub-region; Based on the predicted photovoltaic power generation and the predicted power demand, the predicted storage power of each target sub-area is determined, and based on the predicted storage power, a power dispatch strategy between the multiple target sub-areas is set.

2. The integrated dispatching method for generation, use and storage as claimed in claim 1, characterized in that: The step of obtaining the historical actual weather data corresponding to each meteorological station in the target area and dividing the target area based on the historical actual weather data to obtain multiple target sub-areas specifically includes: Obtaining historical actual weather data corresponding to each weather station in the target area, and calculating the weather index corresponding to the historical actual weather data in each weather station based on a preset weather index calculation formula; The target area is divided into multiple target sub-areas based on the weather index by using the rotational empirical orthogonal decomposition method.

3. The dispatching method for integrated generation, use and storage as claimed in claim 1, characterized in that: The step of setting a power dispatching strategy among the plurality of target sub-areas based on the predicted stored power specifically includes: respectively performing correction processing on the predicted storage power corresponding to each target sub-area to obtain a corrected storage power, and classifying the multiple target sub-areas into supply sub-areas and demand sub-areas based on the corrected storage power; When it is detected that the supply sub-area and the demand sub-area exist at the same time, and the calculated total supply power corresponding to the supply sub-area is greater than the calculated total demand power corresponding to the demand sub-area, respectively calculating the distance between each demand sub-area and all supply sub-areas, and determining the scheduling strategy between the supply sub-area and the demand sub-area based on the distance; When it is detected that the supply sub-area exists and the demand sub-area does not exist, executing a charging strategy for the energy storage system; When it is detected that the demand sub-area exists and the supply sub-area does not exist, a discharging strategy is executed on the energy storage system.

4. The dispatching method for integrated generation, use and storage as claimed in claim 3, characterized in that: The correcting the predicted storage power corresponding to each target sub-area to obtain the corrected storage power specifically includes: Calculate the storage power deviation rate according to multiple historical preset storage powers and the historical actual storage power corresponding to each historical preset storage power, and generate a storage power deviation rate data set; The average storage power deviation rate corresponding to the storage power deviation rate data set is calculated based on the preset confidence, and the predicted storage power corresponding to each target sub-area is corrected based on the average storage power deviation rate to obtain the corrected storage power.

5. The integrated dispatching method for generation, use and storage as claimed in claim 3, characterized in that: The respectively calculating the distances from each demand sub-area to all supply sub-areas, and determining the scheduling strategy between the supply sub-areas and the demand sub-areas based on the distances specifically includes: Obtaining the power requirements corresponding to each of all the demand sub-areas, and sorting all the demand sub-areas based on the power requirements to obtain a demand sub-area sequence; Selecting a single target demand sub-region from the demand sub-region sequence in turn, calculating the distance between the target demand sub-region and all supply sub-regions, and sorting all supply sub-regions in order from near to far based on the distance to obtain a supply sub-region sequence corresponding to the target demand sub-region; The supply sub-area sequence is detected in order from left to right. When it is detected that the target supply power corresponding to the target number of target supply sub-areas is greater than the target demand power corresponding to the target demand sub-area, the target number of target supply sub-areas are used as the power scheduling areas of the target demand sub-areas, and a supply sub-area power scheduling strategy is generated. The supply power corresponding to each of the supply sub-areas is updated based on the supply sub-area power scheduling strategy until all the demand sub-areas in the demand sub-area sequence are selected, and the scheduling strategy between the supply sub-areas and the demand sub-areas is determined.

6. The dispatching method for integrated generation, use and storage as claimed in claim 1, characterized in that: The pre-training process of the first hybrid neural network model specifically includes: Constructing an initial first hybrid neural network model, wherein the first hybrid neural network model includes a first convolutional neural network and a first graph neural network, wherein the first convolutional neural network is connected to the first graph neural network; Acquire power generation data samples and weather data samples corresponding to the same time, and input the weather data samples into the first hybrid neural network model, so that the first convolutional neural network in the first hybrid neural network model respectively extracts features from the weather data samples to obtain a first feature map matrix; Inputting the first feature graph matrix into the first graph neural network based on the first convolutional neural network, so that the first graph neural network performs feature fusion processing on the first feature graph matrix to obtain a first feature fusion graph; Inputting the first feature fusion graph into the first convolutional neural network model based on the first graph neural network, so that the first convolutional neural network performs feature extraction on the first feature fusion graph to obtain predicted power generation data; The predicted power generation data is compared with the power generation data sample to determine a first hybrid neural network model.

7. The dispatching method for integrated generation, use and storage as claimed in claim 6, characterized in that: The step of inputting the weather data samples into the first hybrid neural network model so that the first convolutional neural network in the first hybrid neural network model respectively extracts features from the weather data samples to obtain a first feature graph matrix specifically includes: The weather data samples include irradiance, cloud cover, humidity and average solar sunshine time; Inputting the weather data sample into the first hybrid neural network model, so that the first convolutional neural network in the first hybrid neural network model respectively extracts features of the irradiance, the cloud cover, the humidity and the average solar sunshine time, and respectively obtains feature weights corresponding to the irradiance, the cloud cover, the humidity and the average solar sunshine time; Based on the feature weights, an irradiance feature map matrix, a cloud cover feature map matrix, a humidity feature map matrix and an average solar sunshine time feature map matrix are constructed respectively, and the irradiance feature map matrix, the cloud cover feature map matrix, the humidity feature map matrix and the average solar sunshine time feature map matrix are integrated to obtain a first feature map matrix.

8. A dispatching device integrating generation, use and storage, characterized in that: include: Regional division module, photovoltaic power generation prediction module, power demand prediction module and power dispatch strategy generation module; The area division module is used to obtain the historical actual weather data corresponding to each meteorological station in the target area, and divide the target area based on the historical actual weather data to obtain multiple target sub-areas; The photovoltaic power generation prediction module is used to obtain the historical actual weather data corresponding to each target sub-region, and use the pre-trained first hybrid neural network model to perform data prediction processing on each target sub-region based on the historical actual weather data to obtain the predicted photovoltaic power generation corresponding to each target sub-region; The electricity demand prediction module is used to obtain the historical energy price corresponding to each target sub-area, and use the pre-trained second hybrid neural network model to perform data prediction processing on each target sub-area based on the historical energy price to obtain the predicted electricity demand corresponding to each target sub-area; The power dispatch strategy generation module is used to determine the predicted storage power of each target sub-area based on the predicted photovoltaic power generation and the predicted power demand, and to set the power dispatch strategy between the multiple target sub-areas based on the predicted storage power.

9. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the dispatching method for integrated generation, use and storage as claimed in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the integrated dispatching method for generation, use and storage as described in any one of claims 1 to 7.

Images