Scene decomposition, clustering and evaluation method and system retaining multi-dimensional temporal characteristics
Through seasonal decomposition and self-organizing mapping neural networks, long-term and short-term energy storage needs are decoupled, data dimensions are reduced and reconstructed, the energy storage planning problems in the existing technology are solved, and the retention and evaluation of multi-dimensional timing characteristics are realized. Improved the efficiency and accuracy of energy storage planning.
Patent Information
- Application Number
- CN202411319624.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-23
- Publication Date
- 2024-12-10
- Estimated Expiration
- 2044-09-23
AI Technical Summary
The prior art is difficult to retain the multi-dimensional timing characteristics of data while reducing the difficulty of solving problems, and cannot effectively realize the daytime energy transfer of energy storage, resulting in difficulty in realizing long-term and short-term energy storage planning.
The seasonal decomposition algorithm is used to decouple the long-term energy storage demand scenarios, reduce the dimensions of short-term components based on the self-organized mapping neural network and the K-mean clustering algorithm, and reconstruct the data after decomposition of the dimensionality reduction to establish multi-dimensional timing characteristics. Evaluation indicators evaluate the degree of restoration of reconstructed data.
Through data decomposition, differentiation of trends, seasonality and random fluctuations will be improved, analysis flexibility will be simplified, data complexity will be reduced, storage and computing burden will be reduced, and multi-dimensional timing characteristics will be retained and evaluated, and the efficiency and accuracy of energy storage planning will be improved.
Smart Images
Figure CN118861738B_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the technical field of energy storage data decomposition and clustering, and in particular to a scene decomposition, clustering and evaluation method and system that retains multi-dimensional time series features. Background Art
[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.
[0003] The output of renewable energy sources such as wind turbines and photovoltaics, as well as load demand, fluctuate on a daily and inter-day (weekly, monthly and seasonal) scale due to different factors. For example, the peak photovoltaic output during the day is in excess of demand, the peak load in the evening is in short supply, the load demand is different on weekdays and holidays, the climate affects the fluctuation of wind and solar output, and the seasonal energy imbalance caused by the change of seasons is unbalanced on a long time scale. Among them, short-cycle energy storage (such as batteries, etc.) can be used to balance intra-day fluctuations, but its long-cycle energy storage loss is large and it is not suitable for long-cycle energy storage; long-cycle energy storage (such as hydrogen energy storage, etc.) can realize energy transfer across days, weeks and seasons. How to reasonably plan the long and short-cycle energy storage capacity in the system is a key link in achieving economical and efficient energy use, and the planning problem is inseparable from the processing of scenarios.
[0004] There are two common scenarios for energy storage planning: one is to directly cluster the 8760 hours of data throughout the year to obtain multiple typical days. This method can reduce the dimension of problem solving and speed up the solution, but it can only achieve intra-day energy balance and cannot achieve the daytime energy transfer of energy storage; the other is to directly use the 8760 hours of data throughout the year as the optimization scenario. The dimension of the simulation operation problem of each planning scheme is huge, the solution is difficult, the solution time is long, and the computing power requirements of the computing equipment are high. However, the above two energy storage processing methods are difficult to meet the requirements of long-term and short-term energy storage planning. It is difficult to retain the multi-dimensional time series characteristics of the data while reducing the difficulty of problem solving, and thus it is impossible to ultimately achieve the effective transfer of daytime energy of energy storage. Summary of the invention
[0005] In order to solve the above problems, the present disclosure proposes a scenario decomposition, clustering and evaluation method and system that retains multi-dimensional time series characteristics. The long- and short-cycle energy storage demand scenarios are decoupled through a seasonal decomposition algorithm, and the dimension of the short-cycle component is reduced based on a self-organizing map neural network and a K-means clustering algorithm. The original data is reconstructed using the decomposed and reduced-dimensional data, and a multi-dimensional time series characteristic evaluation index is established to evaluate the degree of restoration of the reconstructed data to the original data information.
[0006] According to some embodiments, the present disclosure adopts the following technical solutions:
[0007] The scene decomposition, clustering and evaluation method retains multi-dimensional temporal features, including:
[0008] Obtain the original net power data of the energy storage system throughout the year;
[0009] Based on the seasonal decomposition algorithm, the net power raw data is decomposed into trend component, period component and residual component, and the trend component is summed up by period as the long period component, and the period component and the residual component are superimposed as the short period component.
[0010] Cut the short-period component according to the decomposition period, calculate the eigenvalue of each period data, and combine the eigenvalue and the short-period component into cluster data;
[0011] Based on the self-organizing map neural network, cluster data is self-organized and clustered to obtain the optimal number of clusters and initial cluster centers. Based on the optimal number of clusters and initial cluster centers, the K-means clustering algorithm is used to perform clustering again to obtain new cluster centers and cluster indexes as typical scenarios for energy storage planning.
[0012] The full time series of short-period components is reconstructed according to the clustering index to obtain the full time series connection data, which is then superimposed with the long-period components to obtain the final reconstructed full-period data. A multi-dimensional time series feature evaluation index is established to quantify the consistency between the reconstructed data and the original data, and comprehensively evaluate the degree of restoration of the reconstructed data to the original data information.
[0013] According to some embodiments, the present disclosure adopts the following technical solutions:
[0014] A scene decomposition, clustering and evaluation system that retains multi-dimensional temporal features, including:
[0015] Data acquisition module, used to obtain the original net power data of the energy storage system throughout the year;
[0016] A decomposition module is used to decompose the net power raw data into a trend component, a period component and a residual component based on a seasonal decomposition algorithm, and sum the trend component by period as a long period component, and superimpose the period component and the residual component as a short period component;
[0017] A feature calculation module is used to cut the short-period component according to the decomposition period, calculate the eigenvalue of each period data, and combine the eigenvalue and the short-period component into cluster data;
[0018] The clustering module is used to perform self-organizing clustering on cluster data based on the self-organizing map neural network to obtain the optimal number of clusters and the initial cluster center; based on the optimal number of clusters and the initial cluster center, the K-means clustering algorithm is used to perform clustering again to obtain new cluster centers and cluster indexes as a typical scenario for energy storage planning;
[0019] The reconstruction and evaluation module is used to reconstruct the full time series of short-period components according to the clustering index, obtain the full time series connection data, and establish multi-dimensional time series feature evaluation indicators to quantify the consistency between the reconstructed data and the original data, and comprehensively evaluate the degree of restoration of the reconstructed data to the original data information.
[0020] According to some embodiments, the present disclosure adopts the following technical solutions:
[0021] A non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the scene decomposition, clustering and evaluation method that retains multi-dimensional temporal features is implemented.
[0022] According to some embodiments, the present disclosure adopts the following technical solutions:
[0023] An electronic device comprises: a processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes the scene decomposition clustering and evaluation method that retains multi-dimensional temporal features.
[0024] Compared with the prior art, the present invention has the following beneficial effects:
[0025] The scene decomposition, clustering and evaluation method for retaining multi-dimensional time series characteristics provided by the present disclosure decomposes the original data into trend component, period component and residual component based on the seasonal decomposition algorithm, sums the trend component by period as the long period component, and superimposes the period component and the residual component as the short period component; the present disclosure can distinguish the trend, seasonality and random fluctuation of data through data decomposition, and improve the flexibility of analysis by analyzing the dynamic changes of data at different time scales, thereby better allocating resources, and at the same time can simplify the complexity of data and reduce the burden of storage and calculation.
[0026] The scene decomposition, clustering and evaluation method for retaining multi-dimensional time series features provided by the present invention divides the short-cycle components according to the decomposition period, calculates the characteristic value of each period data, and uses it together with the short-cycle components as cluster data. The cluster data is self-organized and clustered based on a self-organizing map neural network to determine the optimal number of clusters and initial cluster centers; based on the optimal number of clusters and the initial cluster centers, the K-means clustering algorithm is used to cluster the data again to obtain new cluster centers and cluster indexes. The present invention can automatically determine the optimal number of clusters according to data characteristics, reduce data dimensions while retaining multi-dimensional time series features, and provide a comprehensive evaluation index for the degree of retention of multi-dimensional time series features of reconstructed data.
[0027] The scene decomposition, clustering and evaluation method that retains multi-dimensional time series features provided by the present invention reconstructs the full time series of short-period scenes according to clustering indexes, reconstructs the full-period data by superimposing it with long-period components, and establishes multi-dimensional time series feature evaluation indicators, which can quantify the consistency between the reconstructed data and the original data, thereby comprehensively evaluating the degree of restoration of the reconstructed data to the original data information. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] The accompanying drawings constituting a part of the present disclosure are used to provide a further understanding of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure.
[0029] Figure 1 This is an architecture diagram of a scene decomposition, clustering and evaluation method that retains multi-dimensional temporal features according to an embodiment of the present disclosure;
[0030] Figure 2 A schematic diagram of a self-organizing map neural network according to an embodiment of the present disclosure;
[0031] Figure 3 This is a flow chart of a scene decomposition, clustering and evaluation method that retains multi-dimensional temporal features according to an embodiment of the present disclosure. DETAILED DESCRIPTION
[0032] The present disclosure is further described below in conjunction with the accompanying drawings and embodiments.
[0033] It should be noted that the following detailed descriptions are all illustrative and are intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present disclosure belongs.
[0034] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.
[0035] Example 1
[0036] In one embodiment of the present disclosure, a scene decomposition, clustering and evaluation method that retains multi-dimensional temporal features is provided, including:
[0037] Step 1: Obtain the original net power data of the energy storage system throughout the year;
[0038] Step 2: Based on the seasonal decomposition algorithm, the net power raw data is decomposed into trend component, period component and residual component, and the trend component is summed up by period as the long period component, and the period component and the residual component are superimposed as the short period component;
[0039] Step 3: Cut the short-period component according to the decomposition period, calculate the eigenvalue of each period data, and combine the eigenvalue and the short-period component into cluster data;
[0040] Step 4: Based on the self-organizing map neural network, cluster data is self-organized and clustered to obtain the optimal number of clusters and initial cluster centers. Based on the optimal number of clusters and initial cluster centers, K-means clustering algorithm is used to perform clustering again to obtain new cluster centers and cluster indexes as typical scenarios for energy storage planning.
[0041] Step 5: Reconstruct the full time series of short-period components based on the clustering index to obtain the full time series connection data, and establish a multi-dimensional time series feature evaluation index to quantify the consistency between the reconstructed data and the original data, and comprehensively evaluate the degree of restoration of the reconstructed data to the original data information.
[0042] As an embodiment, in a scenario decomposition, clustering and evaluation method that retains multi-dimensional time series features provided by the present disclosure, a new scenario processing method is proposed, which decouples the long-term and short-term energy storage demand scenarios through a seasonal decomposition algorithm, and further reduces the dimension of the short-term component based on a self-organizing map neural network and a K-means clustering algorithm. At the same time, the original data is reconstructed using the decomposed and reduced-dimensional data. The specific process is as follows:
[0043] Step 1: Obtain the original net power data of the energy storage system throughout the year;
[0044] Specifically, the net power raw data is the net power data for 8760 hours throughout the year. , the net power data is calculated as follows:
[0045] Net power data = Fan output data +Photovoltaic output data - Load demand data (1)
[0046] in, t It is a time index of 8760 hours in a whole year.
[0047] Step 2: Based on the seasonal decomposition algorithm, the net power raw data is decomposed into trend component, period component and residual component, and the trend component is summed up by period as the long period component, and the period component and the residual component are superimposed as the short period component;
[0048] Specifically, the seasonal decomposition algorithm is the STL (Seasonal and Trend decomposition using Loess) decomposition algorithm. Based on the STL scenario decomposition algorithm, the decomposition period is set, and the trend components of each period are summed to obtain the 365-day energy trend component. The energy trend component is the long-period component. The energy trend component is averaged according to the set period to obtain the daily average power trend component, and the short-period component is obtained by subtracting the daily average power trend component from the net power data.
[0049] Furthermore, in the specific STL scene decomposition of the present disclosure, the decomposition period is set to 1 day and 24 hours, then the net power data Decomposition into trend components , periodic component and the residual component ,Right now,
[0050] (2)
[0051] Furthermore, the trend component of each period The sum is the 365-day energy trend component ,
[0052] (3)
[0053] The energy trend component That is the long-period component, and the energy trend component Take the average value according to the set period to obtain the daily average power trend component ,
[0054] (4)
[0055] Among them, the energy trend component and the daily average power trend component As long-term energy data and power data, respectively, and then as net power data Subtract the daily average power trend component Get the short period component ,
[0056] (5)
[0057] in, d A time index for the 365 days of the year.
[0058] Step 3: Cut the short-period component according to the decomposition period, calculate the eigenvalue of each period data, and combine the eigenvalue and the short-period component into cluster data;
[0059] Specifically, the short-period component is cut according to the decomposition period, and the eigenvalues of each period data are calculated. The eigenvalues include: maximum , minimum ,average value , maximum fluctuation value , the eigenvalue and the short-period component after cutting Combine into cluster data , based on the self-organizing map neural network, cluster data is self-organized and clustered to determine the optimal number of clusters and the initial cluster centers .
[0060] Furthermore, the present disclosure sets the decomposition period to 24 hours a day, and cuts the 8760-hour data into a 365-day data matrix, expressed as :
[0061] (6)
[0062] Furthermore, the characteristic values of each period data are calculated, and the characteristic values include the maximum value , minimum ,average value , maximum fluctuation value :
[0063] (7)
[0064] Based on the self-organizing map neural network, clustering data is self-organized and clustered. d The time index for the 365 days of the year, each time a subscript is changed d We get one day's data, cluster the 365 days' data, that is, the clustered data is a 28*365 matrix, and determine the optimal number of clusters and the initial cluster centers The steps are as follows:
[0065] Step 1): Build Layer self-organizing map neural network, and give the competition layer neurons initial weights, set the initial scene number, where the initial scene (initial day number) number is set to .
[0066] Step 2): Input clustering data of different scenarios and obtain the scenarios The corresponding winning neuron ; For the scene Clustering data , calculation and competition for each neuron in the layer The winning neuron is the one with the smallest Euclidean distance . and Now they are respectively and The weight vector of neurons. Set the initial number of iterations .
[0067] (8)
[0068] Step 3): Update the competitive layer neurons based on the neighborhood distance The weight of
[0069] Step 4): Through the neighborhood function and learning rate h kj Update scene; corresponding to the winning neuron Neighborhood The weight vector of the neurons , so that it sends the winning neuron Get closer.
[0070] (9)
[0071] Step 5): Order , repeat the above steps until the error between two adjacent weights is less than a given value, or the upper limit of the number of iterations is reached;
[0072] Step 6): Perform iterative convergence judgment, traverse all scenarios, and output the optimal number of clusters and the initial cluster centers , c 0 is a cluster center:
[0073] (10)
[0074] Furthermore, based on the optimal number of clusters and the initial cluster centers, the K-means clustering algorithm is used to perform clustering again to obtain new cluster centers and cluster indexes;
[0075] In order to further improve the clustering effect, the present invention uses the K-means clustering algorithm to perform the second stage clustering on the clustering data based on the optimal number of clusters and the initial cluster centers obtained by the self-organizing map neural network. c Represents the cluster center index, and gets the new cluster center and clustering index , as a typical scenario for energy storage planning.
[0076] (11)
[0077] (12)
[0078] Step 5: Reconstruct the full time series of the short-period component according to the clustering index to obtain the full time series connection data, and then superimpose the full time series connection data with the long-period component to obtain the final reconstructed full-period data.
[0079] Specifically, the full time series of short-term scenarios is reconstructed based on the cluster index:
[0080] (13)
[0081] (14)
[0082] The full time series connection data and the long period component Perform superposition reconstruction to obtain full cycle data :
[0083] (15)
[0084] in, t It is a 24-hour time index. is the dth day after the cluster center is replaced t Hours of data, It is the data obtained by connecting the data of 8760 hours of 365 days in a year in time series according to the time series relationship, that is, the full time series connection data.
[0085] As an embodiment, the present disclosure also provides a multidimensional time series feature evaluation method, establishes multidimensional time series feature evaluation indicators to evaluate the degree of restoration of reconstructed data to the original data information, and the evaluation indicators selected by the present disclosure include root mean square error, mean absolute percentage error, maximum absolute error, peak percentage error, valley percentage error, determination coefficient, correlation coefficient, slope consistency, daily cycle consistency, single-week cycle consistency and monthly cycle consistency.
[0086] In order to simplify the expression, the original data o Expressed as ,
[0087] Alternative Data r for , 1< i < n , r i The dimension is i Alternative data when is the data dimension, o i When the dimension is i The original data at that time, the indicator calculation formula is as follows:
[0088] The root mean square error (RMSE):
[0089] (16)
[0090] The mean absolute percentage error (MAPE):
[0091] (17)
[0092] The maximum absolute error (MaxAE):
[0093] (18)
[0094] The coefficient of determination:
[0095] (19)
[0096] in is the average value of the original data.
[0097] The correlation coefficient:
[0098] (20)
[0099] in is the mean of the surrogate data.
[0100] The peak percentage error is:
[0101] (twenty one)
[0102] in: is the peak value of the original data, is the index corresponding to the peak value, It is the replacement data under this index.
[0103] The valley percentage error:
[0104] (twenty two)
[0105] in: is the valley value of the original data, is the index corresponding to the valley value, It is the replacement data under this index.
[0106] The slope consistency:
[0107] (twenty three)
[0108] in and are the differences between the original data and the surrogate data, respectively.
[0109] The daily cycle consistency:
[0110] (twenty four)
[0111] (25)
[0112] (26)
[0113] in: and are the daily averages of the original data and the alternative data, respectively.
[0114] The single-week cycle consistency:
[0115] (27)
[0116] (28)
[0117] (29)
[0118] in: and are the weekly averages of the original data and the alternative data, w Indicates a single cycle.
[0119] The monthly cycle consistency:
[0120] (30)
[0121] (31)
[0122] (32)
[0123] in: and are the monthly means of the original data and the alternative data, m Represents the monthly cycle.
[0124] Example 2
[0125] In one embodiment of the present disclosure, a scene decomposition, clustering and evaluation system that retains multi-dimensional temporal features is provided, including:
[0126] Data acquisition module, used to obtain the original net power data of the energy storage system throughout the year;
[0127] A decomposition module is used to decompose the net power raw data into a trend component, a period component and a residual component based on a seasonal decomposition algorithm, and sum the trend component by period as a long period component, and superimpose the period component and the residual component as a short period component;
[0128] A feature calculation module is used to cut the short-period component according to the decomposition period, calculate the eigenvalue of each period data, and combine the eigenvalue and the short-period component into cluster data;
[0129] The clustering module is used to perform self-organizing clustering on cluster data based on the self-organizing map neural network to obtain the optimal number of clusters and the initial cluster center; based on the optimal number of clusters and the initial cluster center, the K-means clustering algorithm is used to perform clustering again to obtain new cluster centers and cluster indexes as a typical scenario for energy storage planning;
[0130] The reconstruction and evaluation module is used to reconstruct the full time series of short-period components according to the clustering index, obtain the full time series connection data, and establish multi-dimensional time series feature evaluation indicators to quantify the consistency between the reconstructed data and the original data, and comprehensively evaluate the degree of restoration of the reconstructed data to the original data information.
[0131] Example 3
[0132] In one embodiment of the present disclosure, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the scene decomposition, clustering and evaluation method that retains multi-dimensional temporal features is implemented.
[0133] Example 4
[0134] In one embodiment, an electronic device is provided, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes the scene decomposition clustering and evaluation method that retains multi-dimensional temporal features.
[0135] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0136] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0137] Although the above describes the specific implementation methods of the present disclosure in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present disclosure. Technical personnel in the relevant field should understand that on the basis of the technical solution of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present disclosure.
Claims
1. A scene decomposition, clustering and evaluation method that retains multi-dimensional temporal features, characterized in that: include: Obtain the original net power data of the energy storage system throughout the year; Based on the seasonal decomposition algorithm, the net power raw data is decomposed into trend component, period component and residual component, and the trend component is summed up by period as the long period component, and the period component and the residual component are superimposed as the short period component. Based on the seasonal decomposition algorithm, the net power raw data is decomposed into a trend component, a cycle component and a residual component, including: based on the seasonal decomposition algorithm, a decomposition cycle is set, the trend component of each cycle is summed to obtain a 365-day energy trend component, the energy trend component is the long-cycle component, the energy trend component is averaged according to the set cycle to obtain a daily average power trend component, and the short-cycle component is obtained by subtracting the daily average power trend component from the net power data; Cut the short-period component according to the decomposition period, calculate the eigenvalue of each period data, and combine the eigenvalue and the short-period component into cluster data; The short-period component is cut according to the decomposition period, and the characteristic value of each periodic data is calculated, including: setting the decomposition period of the short-period component, calculating the characteristic value of each periodic data, the characteristic value includes a maximum value, a minimum value, an average value and a maximum fluctuation value, and combining the maximum value, the minimum value, the average value and the maximum fluctuation value with the short-period component to obtain clustered data; Based on the self-organizing map neural network, cluster data is self-organized and clustered to obtain the optimal number of clusters and initial cluster centers. Based on the optimal number of clusters and initial cluster centers, the K-means clustering algorithm is used to perform clustering again to obtain new cluster centers and cluster indexes as typical scenarios for energy storage planning. Based on the optimal number of clusters and the initial cluster center, the K-means clustering algorithm is used to perform the second stage clustering to obtain a new cluster center. and clustering index , as a typical scenario for energy storage planning, including: in, Represents the cluster center index; represents the optimal number of clusters; Reconstruct the entire time series of short-term scenarios based on clustering indexes: The full time series connection data and the long period component Overlay and reconstruct full cycle data : in, t It is a 24-hour time index. is the dth day after the cluster center is replaced t Hours of data, is the number of clustering iterations, It is the data obtained by connecting the data of 8760 hours of 365 days in a year in time series according to the time series relationship, that is, the full time series connection data; The full time series of short-period components is reconstructed according to the clustering index to obtain the full time series connection data, and a multi-dimensional time series feature evaluation index is established to quantify the consistency between the reconstructed data and the original data, and comprehensively evaluate the degree of restoration of the reconstructed data to the original data information.
2. The scene decomposition, clustering and evaluation method retaining multi-dimensional temporal features according to claim 1, characterized in that: The net power raw data refers to the net power data of 8760 hours throughout the year. The net power data is calculated as follows: wind turbine output data +Photovoltaic output data - Load demand data .
3. The scene decomposition, clustering and evaluation method retaining multi-dimensional temporal features as claimed in claim 1, characterized in that: Based on the self-organizing map neural network, the cluster data is self-organized and clustered to obtain the optimal number of clusters and the initial cluster center, including: Step 1): Construct a self-organizing map neural network, assign initial weights to the competition layer neurons, and set the initial scene number; Step 2): Input clustering data of different scenes and obtain the winning neuron corresponding to the scene; Step 3): Update the weights of the competition layer neurons based on the neighborhood distance; Step 4): Update the scene through the neighborhood function and learning rate; Step 5): Repeat the above steps until the error between two adjacent weights is less than the given value, or the upper limit of the number of iterations is reached; Step 6): After traversing all scenes, output the optimal number of clusters and the initial cluster center.
4. A scene decomposition, clustering and evaluation system that retains multi-dimensional temporal features, specifically implementing a scene decomposition, clustering and evaluation method that retains multi-dimensional temporal features as described in any one of claims 1 to 3, characterized in that: include: Data acquisition module, used to obtain the original net power data of the energy storage system throughout the year; A decomposition module is used to decompose the net power raw data into a trend component, a period component and a residual component based on a seasonal decomposition algorithm, and sum the trend component by period as a long period component, and superimpose the period component and the residual component as a short period component; A feature calculation module is used to cut the short-period component according to the decomposition period, calculate the eigenvalue of each period data, and combine the eigenvalue and the short-period component into cluster data; The clustering module is used to perform self-organizing clustering on cluster data based on the self-organizing map neural network to obtain the optimal number of clusters and the initial cluster center; based on the optimal number of clusters and the initial cluster center, the K-means clustering algorithm is used to perform clustering again to obtain new cluster centers and cluster indexes as a typical scenario for energy storage planning; The reconstruction and evaluation module is used to reconstruct the full time series of short-period components according to the clustering index, obtain the full time series connection data, and establish multi-dimensional time series feature evaluation indicators to quantify the consistency between the reconstructed data and the original data, and comprehensively evaluate the degree of restoration of the reconstructed data to the original data information.
5. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by the processor, the scene decomposition, clustering and evaluation method retaining multi-dimensional temporal features as described in any one of claims 1-3 is implemented.
6. An electronic device, characterized in that: include: A processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes the scene decomposition, clustering and evaluation method that retains multidimensional temporal features as described in any one of claims 1-3.