Distributed energy storage capacity optimal configuration method oriented to source network load storage system

Abstract

The invention relates to the field of energy storage, in particular to a distributed energy storage capacity optimal configuration method oriented to a source network load storage system, which comprises the following steps of: extracting electric power information on the same time characteristic and space characteristic; representing a time sequence dependency relationship of decision behaviors by adopting a long-short-term memory network, and carrying out adaptive learning of a space-time dependency relationship by adopting a graph attention network; taking the battery scale, temperature, cycle index and expansive force variation in the energy storage power station as input conditions to generate a power battery multi-scale mapping model, and finally optimizing the power utilization cost according to the power battery multi-scale mapping model. And constructing a multi-element application model of the energy storage power station, which is based on income cost analysis, operation condition prediction, scheduling instruction response, charging and discharging strategy making, capacity configuration and cross-market quotation decision making, of the energy storage power station. The method has the advantages that power resources can be accurately allocated according to power distribution and load characteristic differences of different regions, and the power dispatching effect is improved.

Description

Technical field

[0001] The present invention relates to the field of energy storage, and in particular to a distributed energy storage capacity optimization configuration method for a source-grid-load-storage-use system. [Background Technology]

[0002] Irrational allocation of distributed energy storage capacity has always been a key and challenging issue in power generation, grid-load-storage systems. In modern power systems, the distribution network, as the key link connecting power production and consumption, is crucial for ensuring the normal functioning of society and the economy. With the continuous adjustment and optimization of the energy structure, the large-scale integration of distributed power sources, and the increasing popularity of energy storage power stations, the operating model of the distribution network is undergoing a profound transformation, shifting from a traditional single-source power supply model to a complex system characterized by the coordinated interaction of power generation, grid-load-storage, and energy storage.

[0003] However, the current charging and discharging characteristics and lifespan of distributed energy storage are intermittent and random. This makes it difficult to predict the charging and discharging characteristics and service life of distributed energy storage, significantly increasing power supply costs. As distribution networks continue to expand and become increasingly complex, centralized dispatch models are struggling to process massive amounts of data and make rapid decisions, making it difficult to meet the demands of coordinated dispatch of sources, grids, loads, and storage. Furthermore, distribution networks in different regions vary significantly in topology, power distribution, and load characteristics. Universal dispatch methods often fail to account for local realities, resulting in poor dispatch results.

[0004] Therefore, a distributed energy storage capacity optimization configuration method for the source-grid-load-storage-use system is proposed. It can comprehensively consider the system operation economy, reliability and other factors, establish an optimization model, determine the optimal capacity and charging and discharging strategy of distributed energy storage, reduce system costs, and improve energy storage utilization efficiency. [Summary of the invention]

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the existing technology and provide a distributed energy storage capacity optimization configuration method for the source-grid-load-storage-use system, which can accurately allocate power resources according to the differences in power supply distribution and load characteristics in different regions, thereby improving the power dispatching effect.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0007] A distributed energy storage capacity optimization configuration method for a source-grid-load-storage-use system includes extracting power information with both temporal and spatial characteristics. The power information includes the charge and discharge power of the energy storage station, state of charge, battery capacity decay, power load level, and temperature.

[0008] A long short-term memory network is used to represent the temporal dependencies of decision-making behaviors. A graph convolutional neural network is used to aggregate the spatial features of power station batteries, and a graph attention network is used for adaptive learning of spatiotemporal dependencies.

[0009] The battery size, temperature, number of cycles, and expansion force changes in the energy storage power station are used as input conditions to generate a multi-scale image model of the power battery. Finally, the electricity cost is optimized based on the multi-scale image model of the power battery.

[0010] Construct a multi-application model for energy storage power stations that makes charging and discharging strategies, capacity allocation, and cross-market quotation decisions based on benefit-cost analysis, operating condition prediction, and dispatch instruction response;

[0011] The formula of the multi-scale image model of power battery is: F(cyc,T,SOC,L)=∫(a1·Δcyc+b1·Δcyc+c1·L)d(Δcyc)·∫(a·ΔT+b)d(ΔT)·f(SOC 2 ), where T is temperature, ΔT is temperature difference, cyc is number of cycles, Δcyc is change in number of cycles, L is battery size, SOC is change in expansion force, a1, b1, and c1 are fitting coefficients of the relationship between the change in expansion force and the number of cycles, which are model parameters or characteristic parameters, and a and b are fitting coefficients of the relationship between the change in expansion force and temperature.

[0012] As a preferred method, the electricity cost is optimized according to the multi-scale image model of the power battery. The calculation formula is as follows:

[0013]

[0014] Where C user is the total cost on the user side, C gen (t) is the power generation cost at time t, C bc (t) is the battery charging cost at time t, C bd (t) is the battery discharge cost at time t, W gen (t) is the power generation load at time t, W bc (t) is the battery charging load at time t, W bd (t) is the battery discharge load at time t.

[0015] Preferably, the wind speed outside the battery, the temperature outside the battery, and the temperature inside the battery are obtained at each time in each area inside the energy storage power station;

[0016] The difference between the temperature inside the battery and the temperature outside the battery at each time in each time period of each area is used as the temperature difference at each time in each time period of each area;

[0017] The temperature difference at each time in each time period of each area and the wind speed outside the battery at the corresponding time are combined into a two-dimensional array.

[0018] Preferably, a cluster analysis algorithm is used to cluster all two-dimensional arrays in each time period of each region to obtain multiple clusters;

[0019] Analyze the differences in battery temperature fluctuations between each area and the remaining adjacent areas during the same time period, and determine the degree of abnormality of each area in each time period;

[0020] The difference in temperature coefficient between each region and the rest of its adjacent regions in the same time period is recorded as the first difference;

[0021] The difference in the abnormality of each region and its adjacent regions in the same time period is recorded as the second difference;

[0022] Compare the first difference to the second difference and record the difference value between them.

[0023] Preferably, the temperature drop index for each time period of each region is determined based on the difference in temperature within the battery between each region and the remaining adjacent regions during the same time period, combined with the regional difference.

[0024] The cooling intensity of each area is determined based on the temperature differences within the battery at all times between different areas and the cooling index.

[0025] Based on the cooling index, the temperature of the energy storage power station is adjusted. The temperature relationship between each element in each cluster and the rest of the elements in the corresponding cluster and the elements in the rest of the clusters is analyzed to determine the intra-cluster and inter-cluster differences of each element in each cluster.

[0026] The ratio of the inter-cluster difference to the intra-cluster difference is recorded as the relative ratio;

[0027] The average of the difference products between each region and all its adjacent regions in each time period is used as the regional difference degree of each region in each time period, and the underlying control adjustment strategy is formed, and the adjustment action is selected based on the strategy goal and algorithm;

[0028] Adjust the operating state of the underlying temperature regulator and calculate the underlying loss difference based on the underlying tuple and loss function.

[0029] Preferably, the parameter combination of the underlying control adjustment strategy is adjusted based on the underlying loss difference, and whether to terminate the operation of the underlying control adjustment strategy is determined based on the number of times the parameter combination is adjusted;

[0030] Sort the bottom layer loss differences to obtain the minimum bottom layer loss difference, and mark the bottom layer temperature in the bottom layer tuple corresponding to the minimum bottom layer loss value as the real-time ambient temperature data;

[0031] Based on the real-time ambient temperature data and the target temperature, it is determined whether the temperature adjustment of the energy storage power station is completed. If the real-time ambient temperature data is equal to the target temperature, it is determined that the temperature adjustment of the energy storage power station is completed and the adjustment is ended. If the real-time ambient temperature data is not equal to the target temperature, it is determined that the temperature adjustment of the energy storage power station is not completed.

[0032] Preferably, the graph neural network algorithm and the long short-term memory network algorithm are used to train and learn the multi-application model of the energy storage power station to obtain a spatiotemporal causal correlation matrix representing the correlation characteristics between the energy storage power station and other batteries in the power system. The graph convolutional neural network is used to extract the spatial correlation characteristics of the energy storage power station nodes in the sample data set of the multi-application model of the energy storage power station. The graph convolutional neural network obtains the high-order correlation pattern of the energy storage power station in the power grid topology structure by aggregating the feature information of the energy storage power station nodes and their neighboring nodes layer by layer.

[0033] Preferably, a long short-term memory network is used to extract the dependency of the spatial correlation features of the energy storage power station nodes output by the graph convolutional neural network in the time dimension to obtain the output result of the long short-term memory network;

[0034] The output results of the long short-term memory network are mapped to the probability distribution of the multiple application modes of the energy storage power station through a fully connected layer. The loss function for the classification of energy storage power station nodes is constructed. The model parameters of the graph convolutional neural network and the long short-term memory network are jointly optimized in an end-to-end manner to train the graph convolutional neural network and the long short-term memory network model.

[0035] A distributed energy storage capacity optimization configuration system for a source-grid-load-storage-use system includes an extraction module for extracting power information with both temporal and spatial characteristics. The power information includes the charge and discharge power of the energy storage station, state of charge, battery capacity decay, power load level, and temperature.

[0036] The convolutional neural network module uses a long short-term memory network to represent the temporal dependencies of decision-making behaviors. It uses a graph convolutional neural network to aggregate the spatial features of power station batteries, and a graph attention network for adaptive learning of spatiotemporal dependencies.

[0037] The battery multi-scale imaging module is used to generate a multi-scale imaging model of the power battery using the battery size, temperature, number of cycles, and expansion force changes in the energy storage power station as input conditions. The module ultimately optimizes electricity costs based on the multi-scale imaging model of the power battery.

[0038] The operation module constructs a multi-application model for energy storage power stations, which makes charging and discharging strategies, capacity configuration, and cross-market quotation decisions based on benefit-cost analysis, operating condition prediction, and dispatch instruction response.

[0039] A computer device includes a processor and a memory, wherein the processor is connected to the memory, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the computer device performs at least one step in the above method.

[0040] The beneficial effects of the present invention are as follows:

[0041] 1. The formula of the multi-scale image model of power battery is: F(cyc,T,SOC,L)=∫(a1·Δcyc+b1·Δcyc+c1·L)d(Δcyc)·∫(a·ΔT+b)d(ΔT)·f(SOC 2 ), where T is temperature, ΔT is temperature difference, cyc is number of cycles, Δcyc is change in number of cycles, L is battery size, SOC is change in expansion force, a1, b1, and c1 are fitting coefficients of the relationship between the change in expansion force and the number of cycles, which are model parameters or characteristic parameters, and a and b are fitting coefficients of the relationship between the change in expansion force and temperature.

[0042] The electricity cost is optimized based on the multi-scale image model of the power battery. The calculation formula is as follows:

[0043]

[0044] W gen (t) = nF;

[0045] W bd (t) = mF;

[0046] Where C user is the total cost on the user side, C gen (t) is the power generation cost at time t, C bc (t) is the battery charging cost at time t, C bd (t) is the battery discharge cost at time t, W gen (t) is the power generation load at time t, W bc (t) is the battery charging load at time t, W bd (t) is the battery discharge load at time t, and n and m are characteristic parameters.

[0047] 2. Using cluster analysis algorithms, cluster all two-dimensional arrays in each time period of each region to obtain multiple clusters;

[0048] Analyze the differences in battery temperature fluctuations between each area and the remaining adjacent areas during the same time period, and determine the degree of abnormality of each area in each time period;

[0049] The difference in temperature coefficient between each region and the rest of its adjacent regions in the same time period is recorded as the first difference;

[0050] The difference in the abnormality of each region and its adjacent regions in the same time period is recorded as the second difference;

[0051] Compare the first difference to the second difference and record the difference value between them.

[0052] It should be noted that the larger the first difference, the more the temperature state between adjacent areas is affected by external weather conditions. The larger the second difference, the greater the difference in the random changes in the air affecting temperature over a short period of time between adjacent areas. The larger the regional difference, the more inconsistent the temperature state of each area is with its neighbors. Random changes in the air affecting temperature refer to factors such as wind speed, dust, and humidity.

[0053] 3. Based on the cooling index, the temperature of the energy storage power station is adjusted. The temperature relationship between each element in each cluster and the rest of the elements in the corresponding cluster and the elements in the rest of the clusters is analyzed to determine the intra-cluster and inter-cluster differences of each element in each cluster.

[0054] The ratio of the inter-cluster difference to the intra-cluster difference is recorded as the relative ratio;

[0055] The average of the difference products between each region and all its adjacent regions in each time period is used as the regional difference degree of each region in each time period, and the underlying control adjustment strategy is formed, and the adjustment action is selected based on the strategy goal and algorithm;

[0056] Adjust the operating status of the underlying temperature regulator and calculate the underlying loss difference based on the underlying tuple and loss function. The calculation formula is as follows: Where mse is the loss function, is the input value corresponding to the randomly selected battery tuple for the i-th time, x is the input value corresponding to all battery tuples as a whole, and n is the number of battery tuples.

[0057] These features and advantages of the present invention will be disclosed in detail in the following specific embodiments and drawings.

Brief Description of the Drawings

[0058] The present invention will be further described below with reference to the accompanying drawings:

[0059] Figure 1 Schematic diagram of the flow of the energy storage capacity optimization configuration method according to the first embodiment of the present invention. [Specific implementation method]

[0060] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0061] The following first describes the concepts involved in this application with reference to the accompanying drawings. It should be noted that the following description of each concept is only for the purpose of making the content of this application easier to understand and does not limit the scope of protection of this application.

[0062] Example 1:

[0063] A distributed energy storage capacity optimization configuration method for source-grid-load-storage-use system, such as Figure 1 As shown, it includes extracting power information based on the same time characteristics and spatial characteristics. The power information includes the charging and discharging power of the energy storage station, the state of charge, the battery capacity attenuation state, the power load level, and the temperature.

[0064] A long short-term memory network is used to represent the temporal dependencies of decision-making behaviors. A graph convolutional neural network is used to aggregate the spatial features of power station batteries, and a graph attention network is used for adaptive learning of spatiotemporal dependencies.

[0065] The battery size, temperature, number of cycles, and change in expansion force in the energy storage power station are used as input conditions to generate a multi-scale image model of the power battery. Ultimately, the electricity cost is optimized based on the multi-scale image model. The amount of energy stored is related to the battery capacity. During the battery charging and discharging process, the battery expansion volume changes with the lithium ion concentration, that is, there is a correlation between the battery expansion volume and the battery state of charge (SOC) and the internal and external temperatures. However, due to the non-monotonic phase transition of the positive electrode material structure, the expansion force and battery state of charge (SOC) of lithium iron phosphate batteries are nonlinear, which poses a significant challenge for applications in the field of SOC estimation.

[0066] The present invention plans the energy storage capacity of the battery based on the two concepts of expansion force variation and temperature, thereby improving the environmental adaptability and portability of the model.

[0067] Constructing a multi-application model for an energy storage power station that makes charging and discharging strategies, capacity configuration, and cross-market quotation decisions based on benefit-cost analysis, operating condition prediction, and dispatch instruction response. The multi-application model for an energy storage power station that makes cross-market quotation decisions may also include: constructing a sample data set for the multi-application model for an energy storage power station, where each sample in the sample data set for the multi-application model for an energy storage power station consists of a time series matrix of attributes of the energy storage power station and its related power battery nodes, an adjacency matrix of the grid topology, and a target application mode of the energy storage power station at each moment;

[0068] The formula of the multi-scale image model of power battery is: F(cyc,T,SOC,L)=∫(a1·Δcyc+b1·Δcyc+c1·L)d(Δcyc)·∫(a·ΔT+b)d(ΔT)·f(SOC 2 ), where T is temperature, ΔT is temperature difference, cyc is number of cycles, Δcyc is change in number of cycles, L is battery size, SOC is change in expansion force, a1, b1, and c1 are fitting coefficients of the relationship between the change in expansion force and the number of cycles, which are model parameters or characteristic parameters, and a and b are fitting coefficients of the relationship between the change in expansion force and temperature.

[0069] The electricity cost is optimized based on the multi-scale image model of the power battery. The calculation formula is as follows:

[0070]

[0071] W gen (t) = nF;

[0072] W bd (t) = mF;

[0073] Where C user is the total cost on the user side, C gen (t) is the power generation cost at time t, C bc (t) is the battery charging cost at time t, C bd (t) is the battery discharge cost at time t, W gen (t) is the power generation load at time t, W bc (t) is the battery charging load at time t, W bd (t) is the battery discharge load at time t, and n and m are characteristic parameters.

[0074] It is worth mentioning that the battery charging load and battery discharging load are related to the operating temperature and SOC of the energy storage power station battery.

[0075] Obtain the wind speed outside the battery, the temperature outside the battery, and the temperature inside the battery at each time in each area inside the energy storage power station;

[0076] The difference between the temperature inside the battery and the temperature outside the battery at each time in each time period of each area is used as the temperature difference at each time in each time period of each area;

[0077] The temperature difference at each time in each time period of each area and the wind speed outside the battery at the corresponding time are combined into a two-dimensional array.

[0078] Using cluster analysis algorithm, all two-dimensional arrays in each time period of each region are clustered to obtain multiple clusters;

[0079] Analyze the differences in battery temperature fluctuations between each area and the remaining adjacent areas during the same time period, and determine the degree of abnormality of each area in each time period;

[0080] The difference in temperature coefficient between each region and the rest of its adjacent regions in the same time period is recorded as the first difference;

[0081] The difference in the abnormality of each region and its adjacent regions in the same time period is recorded as the second difference;

[0082] Compare the first difference to the second difference and record the difference value between them.

[0083] It should be noted that the larger the first difference, the more the temperature state between adjacent areas is affected by external weather conditions. The larger the second difference, the greater the difference in the random changes in the air affecting temperature over a short period of time between adjacent areas. The larger the regional difference, the more inconsistent the temperature state of each area is with its neighbors. Random changes in the air affecting temperature refer to factors such as wind speed, dust, and humidity.

[0084] Example 2:

[0085] Determine the cooling index for each time period of each area based on the difference in battery temperature between each area and the remaining adjacent areas during the same time period, combined with the regional differences;

[0086] The cooling intensity of each area is determined based on the temperature differences within the battery at all times between different areas and the cooling index.

[0087] Based on the cooling index, the temperature of the energy storage power station is adjusted. The temperature relationship between each element in each cluster and the rest of the elements in the corresponding cluster and the elements in the rest of the clusters is analyzed to determine the intra-cluster and inter-cluster differences of each element in each cluster.

[0088] The ratio of the inter-cluster difference to the intra-cluster difference is recorded as the relative ratio;

[0089] The average of the difference products between each region and all its adjacent regions in each time period is used as the regional difference degree of each region in each time period, and the underlying control adjustment strategy is formed, and the adjustment action is selected based on the strategy goal and algorithm;

[0090] Adjust the operating status of the underlying temperature regulator and calculate the underlying loss difference based on the underlying tuple and loss function. The calculation formula is as follows: Where mse is the loss function, is the input value corresponding to the randomly selected battery tuple for the i-th time, x is the input value corresponding to all battery tuples as a whole, and n is the number of battery tuples.

[0091] Adjust the parameter combination of the underlying control adjustment strategy based on the underlying loss difference, and determine whether to terminate the operation of the underlying control adjustment strategy based on the number of times the parameter combination is adjusted;

[0092] Sort the bottom layer loss differences to obtain the minimum bottom layer loss difference, and mark the bottom layer temperature in the bottom layer tuple corresponding to the minimum bottom layer loss value as the real-time ambient temperature data;

[0093] Based on the real-time ambient temperature data and the target temperature, it is determined whether the temperature adjustment of the energy storage power station is completed. If the real-time ambient temperature data is equal to the target temperature, it is determined that the temperature adjustment of the energy storage power station is completed and the adjustment is ended. If the real-time ambient temperature data is not equal to the target temperature, it is determined that the temperature adjustment of the energy storage power station is not completed.

[0094] When it is determined that the external ambient temperature needs to be adjusted, the target temperature is selected based on the environmental status data and the top-level target temperature strategy to obtain a target temperature suitable for the current environment. The target temperature can then be set as the policy target of the corresponding bottom-level control and adjustment strategy to control the ambient temperature, so that the ambient temperature can be quickly increased or decreased, and can change with changes in the complex environment, thereby achieving a stable relative relationship between the ambient temperature and the complex environment, and significantly improving the temperature controllability of the temperature control method;

[0095] Secondly, to ensure the effectiveness of temperature regulation, the system determines whether the ambient temperature adjustment is complete based on the real-time ambient temperature data and the target temperature during the temperature adjustment process. If it is not complete, that is, if the real-time ambient temperature data is not equal to the target temperature, the system can continue to adjust the ambient temperature based on the target temperature and the underlying control and adjustment strategy until the real-time ambient temperature data reaches the target temperature, thus avoiding wasting temperature adjustment resources after the ambient temperature adjustment is completed. Similarly, the internal temperature of the energy storage power station can also be determined and adjusted according to this judgment method.

[0096] The graph neural network algorithm and long short-term memory network algorithm are used to train and learn the multi-application model of energy storage power stations, and a spatiotemporal causal correlation matrix representing the correlation characteristics between energy storage power stations and other batteries in the power system is obtained. The graph convolutional neural network is used to extract the spatial correlation characteristics of energy storage power station nodes in the sample data set of the multi-application model of energy storage power stations. The graph convolutional neural network obtains the high-order correlation pattern of energy storage power stations in the power grid topology structure by aggregating the feature information of energy storage power station nodes and their neighboring nodes layer by layer.

[0097] A long short-term memory network is used to extract the temporal dependency of the spatial correlation features of energy storage power station nodes output by the graph convolutional neural network, and the output of the long short-term memory network is obtained.

[0098] The output results of the long short-term memory network are mapped to the probability distribution of the multiple application modes of the energy storage power station through a fully connected layer. The loss function for the classification of energy storage power station nodes is constructed. The model parameters of the graph convolutional neural network and the long short-term memory network are jointly optimized in an end-to-end manner to train the graph convolutional neural network and the long short-term memory network model.

[0099] For example, we can perform multi-level recursive convolution processing on the real-time data of the ambient temperature of the energy storage power station to extract the spatiotemporal features. The calculation formula is as follows:

[0100]

[0101] in, is the spatiotemporal convolution feature of the lth convolution layer at time t, is the activation function, is the convolution kernel, x' f (t,f) is the real-time data of ambient temperature after noise filtering, is the bias term of the lth convolutional layer at time t, is the feature weight coefficient of the lth convolutional layer at time t to the mth convolutional layer at time t-1, is the number of convolutional layers, is the spatiotemporal convolution feature of the mth convolutional layer at time t-1.

[0102] By processing the data enhancement algorithm based on a multi-level recursive convolutional network, real-time ambient temperature data with enhanced stability is obtained.

[0103] Through the graph neural network algorithm and the long short-term memory network algorithm, the graph convolutional neural network aggregates the characteristic information of the energy storage power station node and its neighboring nodes layer by layer to obtain the high-order correlation pattern of the energy storage power station in the power grid topology structure, thereby improving the rationality and accuracy of the electricity economic benefit evaluation. The algorithm is relatively simple and has low complexity. Verification shows that when obtaining an average error similar to that of the above embodiment, only about 30% of the collected data volume is required, reducing the workload of collecting electricity efficiency data, effectively expanding the applicable working conditions of the system, reducing the time cost of data collection, and improving the rationality of the system.

[0104] A distributed energy storage capacity optimization configuration system for a source-grid-load-storage-use system includes an extraction module for extracting power information with both temporal and spatial characteristics. The power information includes the charge and discharge power of the energy storage station, state of charge, battery capacity decay, power load level, and temperature.

[0105] The convolutional neural network module uses a long short-term memory network to represent the temporal dependencies of decision-making behaviors. It uses a graph convolutional neural network to aggregate the spatial features of power station batteries, and a graph attention network for adaptive learning of spatiotemporal dependencies.

[0106] The battery multi-scale imaging module is used to generate a multi-scale imaging model of the power battery using the battery size, temperature, number of cycles, and expansion force changes in the energy storage power station as input conditions. The module ultimately optimizes electricity costs based on the multi-scale imaging model of the power battery.

[0107] The operation module constructs a multi-application model for energy storage power stations, which makes charging and discharging strategies, capacity configuration, and cross-market quotation decisions based on benefit-cost analysis, operating condition prediction, and dispatch instruction response.

[0108] A computer device includes a processor and a memory, wherein the processor is connected to the memory, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the computer device performs at least one step in the above method.

[0109] The above are only specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Those skilled in the art should understand that the present invention includes but is not limited to the contents described in the drawings and the above specific embodiments. Any modifications that do not deviate from the functional and structural principles of the present invention are intended to be included within the scope of the claims.

Claims

1. A distributed energy storage capacity optimization configuration method for a source-grid-load-storage-use system, characterized by: This includes extracting power information based on both temporal and spatial characteristics, including the charging and discharging power of the energy storage station, state of charge, battery capacity attenuation, power load level, and temperature. A long short-term memory network is used to represent the temporal dependencies of decision-making behaviors. A graph convolutional neural network is used to aggregate the spatial features of power station batteries, and a graph attention network is used for adaptive learning of spatiotemporal dependencies. The battery size, temperature, number of cycles, and expansion force changes in the energy storage power station are used as input conditions to generate a multi-scale image model of the power battery. Finally, the electricity cost is optimized based on the multi-scale image model of the power battery. Construct a multi-application model for energy storage power stations that makes charging and discharging strategies, capacity allocation, and cross-market quotation decisions based on benefit-cost analysis, operating condition prediction, and dispatch instruction response; The formula of the multi-scale image model of the power battery is: F(cyc,T,SOC,L)=∫(a1·Δcyc+b1·Δcyc+c1·L)d(Δcyc)·∫(a·ΔT+b)d(ΔT)·f(SOC 2 ), where T is temperature, ΔT is temperature difference, cyc is number of cycles, Δcyc is change in number of cycles, L is battery size, SOC is change in expansion force, a1, b1, and c1 are fitting coefficients of the relationship between the change in expansion force and the number of cycles, which are model parameters or characteristic parameters, and a and b are fitting coefficients of the relationship between the change in expansion force and temperature.

2. The method for optimizing the configuration of distributed energy storage capacity for a source-grid-load-storage-use system according to claim 1, characterized in that: The electricity cost is optimized based on the multi-scale image model of the power battery. The calculation formula is as follows: Where C user is the total cost on the user side, C gen (t) is the power generation cost at time t, C bc (t) is the battery charging cost at time t, C bd (t) is the battery discharge cost at time t, W gen (t) is the power generation load at time t, W bc (t) is the battery charging load at time t, W bd (t) is the battery discharge load at time t.

3. The method for optimizing the configuration of distributed energy storage capacity for a source-grid-load-storage-use system according to claim 1, characterized in that: Obtain the wind speed outside the battery, the temperature outside the battery, and the temperature inside the battery at each time in each area inside the energy storage power station; The difference between the temperature inside the battery and the temperature outside the battery at each time in each time period of each area is used as the temperature difference at each time in each time period of each area; The temperature difference at each time in each time period of each area and the wind speed outside the battery at the corresponding time are combined into a two-dimensional array.

4. The method for optimizing the configuration of distributed energy storage capacity for a source-grid-load-storage-use system according to claim 3, characterized in that: Using cluster analysis algorithm, all two-dimensional arrays in each time period of each region are clustered to obtain multiple clusters; Analyze the differences in battery temperature fluctuations between each area and the remaining adjacent areas during the same time period, and determine the degree of abnormality of each area in each time period; The difference in the temperature coefficient between each region and the rest of the adjacent regions in the same time period is recorded as a first difference; The difference in the abnormality of each region and its adjacent regions in the same time period is recorded as the second difference; Compare the first difference to the second difference and record the difference value between them.

5. The method for optimizing the configuration of distributed energy storage capacity for a source-grid-load-storage-use system according to claim 1, characterized in that: Determine a cooling index for each time period of each area based on the difference in battery temperature between each area and the remaining adjacent areas during the same time period, combined with the regional differences; The cooling intensity of each area is determined based on the temperature differences within the battery at all times between different areas and the cooling index. Based on the cooling index, the temperature of the energy storage power station is adjusted, and the temperature relationship between each element in each cluster and the remaining elements in the corresponding cluster and the elements in the remaining clusters is analyzed to determine the intra-cluster difference and inter-cluster difference of each element in each cluster; The ratio of the inter-cluster difference to the intra-cluster difference is recorded as the relative ratio; The average of the difference products between each region and all its adjacent regions in each time period is used as the regional difference degree of each region in each time period, and an underlying control adjustment strategy is formed, and an adjustment action is selected based on the strategy goal and algorithm; Adjust the operating state of the underlying temperature regulator and calculate the underlying loss difference based on the underlying tuple and loss function.

6. The method for optimizing the configuration of distributed energy storage capacity for a source-grid-load-storage-use system according to claim 5, characterized in that: Adjusting the parameter combination of the underlying control adjustment strategy based on the underlying loss difference, and determining whether to terminate the operation of the underlying control adjustment strategy based on the number of times the parameter combination is adjusted; Sort the bottom layer loss differences to obtain the minimum bottom layer loss difference, and mark the bottom layer temperature in the bottom layer tuple corresponding to the minimum bottom layer loss value as the real-time ambient temperature data; Based on the real-time ambient temperature data and the target temperature, it is determined whether the temperature adjustment of the energy storage power station is completed. If the real-time ambient temperature data is equal to the target temperature, it is determined that the temperature adjustment of the energy storage power station is completed and the adjustment is ended. If the real-time ambient temperature data is not equal to the target temperature, it is determined that the temperature adjustment of the energy storage power station is not completed.

7. The method for optimizing the configuration of distributed energy storage capacity for a source-grid-load-storage-use system according to claim 1, characterized in that: The multi-application model of the energy storage power station is trained and learned using a graph neural network algorithm and a long short-term memory network algorithm to obtain a spatiotemporal causal association matrix representing the correlation characteristics between the energy storage power station and other batteries in the power system. A graph convolutional neural network is used to extract the spatial correlation characteristics of the energy storage power station nodes in the sample data set of the multi-application model of the energy storage power station. The graph convolutional neural network obtains the high-order correlation pattern of the energy storage power station in the power grid topology structure by aggregating the feature information of the energy storage power station node and its neighboring nodes layer by layer.

8. The method for optimizing the configuration of distributed energy storage capacity for a source-grid-load-storage-use system according to claim 7, characterized in that: A long short-term memory network is used to extract the dependency of the spatial correlation features of the energy storage power station nodes output by the graph convolutional neural network in the time dimension, and an output result of the long short-term memory network is obtained; The output results of the long short-term memory network are mapped to the probability distribution of the multi-application modes of the energy storage power station through a fully connected layer, and a loss function for the classification of energy storage power station nodes is constructed. The model parameters of the graph convolutional neural network and the long short-term memory network are jointly optimized in an end-to-end manner to train the graph convolutional neural network and long short-term memory network models.

9. A distributed energy storage capacity optimization configuration system for source-grid-load-storage-use systems, characterized by: It includes an extraction module for extracting power information based on the same time characteristics and spatial characteristics, wherein the power information includes the charging and discharging power of the energy storage station, the state of charge, the battery capacity attenuation state, the power load level, and the temperature; The convolutional neural network module uses a long short-term memory network to represent the temporal dependencies of decision-making behaviors. It uses a graph convolutional neural network to aggregate the spatial features of power station batteries, and a graph attention network for adaptive learning of spatiotemporal dependencies. The battery multi-scale imaging module is used to generate a multi-scale imaging model of the power battery using the battery size, temperature, number of cycles, and expansion force changes in the energy storage power station as input conditions. The module ultimately optimizes electricity costs based on the multi-scale imaging model of the power battery. The operation module constructs a multi-application model for energy storage power stations, which makes charging and discharging strategies, capacity configuration, and cross-market quotation decisions based on benefit-cost analysis, operating condition prediction, and dispatch instruction response.

10. A computer device comprising a processor and a memory, wherein the processor is connected to the memory, and the memory is used to store a computer program, wherein: The processor is configured to execute the computer program stored in the memory, so that the computer device performs at least one step of the method according to claims 1 to 8.

Images