Method and device for identifying influence factors of power grid receiving capacity under multi-form load

Abstract

The application belongs to the technical field of electric power, and provides a method and device for identifying influencing factors of power grid receiving capacity under multi-form load. Firstly, multi-scale data such as annual, monthly and daily data of industrial electricity, agricultural electricity, transportation electricity, municipal life electricity and new energy electricity generation are obtained. Secondly, the obtained power load data is subjected to feature extraction and cluster analysis by means of improved path integral method and multi-dimensional extended space method. Thirdly, the correlation between the power grid receiving capacity and the cluster feature data is analyzed by means of improved Spearman method. Finally, the running performance under the access situation of multiple loads is predicted by means of improved LSTM algorithm, so as to identify the influencing factors of the power grid receiving capacity under the access of multi-form load. The method for identifying the influencing factors of the power grid receiving capacity under the access of multi-form load systematically considers the multi-aspect influence of different load types on the power grid.

Description

Technical Field

[0001] The present invention belongs to the field of electric power technology, and in particular relates to a method and device for identifying factors influencing the acceptance capacity of a power grid under multi-form loads. Background Art

[0002] With the growing demand for electricity and the rapid development of new loads, traditional power grids are facing increasing pressure. This is especially true when diverse loads (such as industrial loads, renewable energy loads, and electric vehicle charging loads) and renewable energy sources (such as wind power and photovoltaics) are connected to the grid, posing new challenges to the grid's capacity and stability. The varying characteristics, volatility, and uncertainty of these different load types and renewable energy sources make the grid susceptible to voltage fluctuations, frequency instability, and power quality degradation when these loads and renewable energy sources are connected. Furthermore, the intermittent and random nature of renewable energy sources can lead to variability in power flows, reducing the predictability of grid dispatch. Large-scale integration of renewable energy sources can also affect system frequency stability. To ensure the safe and stable operation of the power grid, a comprehensive and systematic analysis of the impact of different loads and renewable energy on the grid is essential. Summary of the Invention

[0003] To solve the above technical problems, the present invention provides a method and device for identifying factors affecting the acceptance capacity of a power grid under multi-mode loads to solve the problems in the prior art. The technical solution adopted by the present invention is:

[0004] The method for identifying factors affecting the grid's acceptance capacity under multi-form loads includes:

[0005] Step 1: Obtain various types of multi-scale load data;

[0006] Step 2: Extract various types of multi-scale load data features by the path multi-integral method;

[0007] Step 3: Cluster analysis of multi-scale load data features based on multi-dimensional extended space learning method;

[0008] Step 4: Use the improved Spearson method to analyze the correlation between the grid acceptance capacity and the characteristics of the clustered multi-scale load data;

[0009] Step 5: Use the improved LSTM algorithm to predict the grid's operating performance under various load access scenarios;

[0010] Wherein, step 2 includes:

[0011] The various types of multi-scale load data obtained in step 1 are connected in chronological order as a special case of the path. The data are reconstructed and expanded to the r-dimensional path space to obtain a continuous mapping X. The coordinates of the r-dimensional path at time t are obtained as follows:

[0012]

[0013] in: is the coordinate of the r-dimensional path at time t, r is the dimension, and t is the time series of the data;

[0014] Convert the superscripts of the mapping X into a sequence

[0015] i1,i2,...,i k ∈{1,...,r}

[0016] Where: i k is the kth load data, k is the total number of load types;

[0017] The one-dimensional k-fold integral is expressed as:

[0018]

[0019] Where: T is the total time, F() is the integral function of each type of multi-scale load data, and d is the differential symbol;

[0020] For different types of loads, the r-dimensional multi-integral is expressed as:

[0021]

[0022] The solution is:

[0023]

[0024] in: is the multi-scale load data feature; r is the dimension of the data, such as converting one-dimensional data into high-dimensional data features such as three-dimensional space vectors; k is the type of load data, including industrial load, renewable energy load, electric vehicle charging load, etc.

[0025] Furthermore, step 3 includes:

[0026] For multi-scale load data characteristics set up To add new data, and P k Projected to In the r-dimensional subspace formed, the cluster labels are obtained, and the objective function is:

[0027]

[0028] in: For the i-th newly added series of the k-th load type, is the multi-scale load data feature; is the optimal projection vector of the kth load, It constrains the self-representation matrix for various matrix norms, and γ1 and γ2 are trade-off parameters;

[0029] By minimizing the norm, the projection between historical data and new data is established. The optimal projection vector is:

[0030]

[0031] Where: E is a unit vector;

[0032] Obtain and The residual of Zhang Cheng's subspace in the jth dimension is:

[0033]

[0034] in: for and The residual of the spanned subspace in the jth dimension; is the jth non-zero element in the optimal vector;

[0035] According to the above residuals, by comparing The residual value in each dimension determines the minimum residual dimension as the cluster label:

[0036]

[0037] in: is the cluster label of the data, arg(*) is the complex argument principal value function;

[0038] Get the multi-scale load data characteristics after clustering for:

[0039]

[0040] Furthermore, step 4 includes:

[0041] The multi-scale load data features after clustering and grid capacity data Find the mean of each:

[0042]

[0043] in: is the mean value of the influencing factor data, is the mean value of the grid acceptance capacity data;

[0044] Find the population covariance based on the mean:

[0045]

[0046] in: is the covariance of influencing factors and grid acceptance capacity data;

[0047] Calculate the correlation coefficient of the two types of data respectively:

[0048]

[0049] in: and are the variances of influencing factors and grid acceptance capacity data respectively;

[0050] The Pearson correlation coefficient is obtained by calculation:

[0051]

[0052] in: is the Pearson correlation coefficient between the influencing factors and the grid acceptance capacity data;

[0053] Convert the clustered multi-scale load data into hierarchical data The Spearman coefficient is obtained by calculating the correlation coefficient of the rank data:

[0054]

[0055] in: is the Spearman correlation coefficient between the influencing factors and the grid acceptance capacity data;

[0056] By using the Pearson coefficient and the Spearman coefficient, the improved Spearman correlation coefficient is obtained:

[0057]

[0058] in: is the multi-scale load data feature after clustering, is the data of the maximum impact that the power grid can withstand, α1 is the proportional coefficient of Pearson correlation, and α2 is the proportional coefficient of improved Spearman correlation; is the correlation coefficient obtained by the Pearson correlation method, is the correlation coefficient obtained by the Spearman correlation method.

[0059] Furthermore, step 5 includes:

[0060] Multi-scale load data features Conduct static graph learning and dynamic graph learning:

[0061]

[0062] in: A H ,A Gare learnable parameters, H and G are query parameters and path parameters in the attention mechanism respectively, and L s is the adjacency matrix of the static graph. The ReLU function is used to eliminate negative connections in the matrix. The Softmax function normalizes each row of the matrix to preserve the relative relationship of the data:

[0063]

[0064] Where: ∥*∥ is the Euclidean norm; DistNet(*) is the distance network; L d is the adjacency matrix of the dynamic graph;

[0065] Through convolution operations, the clustered multi-scale load data features are combined with the multi-scale load data features learned through static and dynamic graphs:

[0066]

[0067] in: and is the data of the feature sequence after static graph and dynamic graph learning; β1 and β2 are hyperparameters; W1 and W2 are learnable parameters, and b1 and b2 are learning biases;

[0068] The prediction of the operating status of the power grid is:

[0069]

[0070] Where: W′1 and W′2 are learnable parameters of the fully connected layer, b′1 and b′2 are learnable biases, is the predicted value of the power grid operation status.

[0071] The device for identifying factors affecting the acceptance capacity of a power grid under multi-form loads includes: a feature extraction module, a feature clustering module, a correlation analysis module and a prediction module; wherein:

[0072] The feature extraction module is connected to the power grid to obtain various types of multi-scale load data, and extracts the features of various types of multi-scale load data through the path multi-integral method;

[0073] The feature clustering module performs cluster analysis on multi-scale load data features based on the multi-dimensional extended space learning method;

[0074] The correlation analysis module uses the improved Spearson method to analyze the correlation between the grid acceptance capacity and the multi-scale load data characteristics after clustering;

[0075] The prediction module uses an improved LSTM algorithm to predict the operating performance of the power grid under various load access scenarios;

[0076] A computer-readable storage medium stores a computer program. When the computer program is run on a computer, the computer is enabled to execute a method for identifying factors affecting the acceptance capacity of a power grid under multi-modal loads.

[0077] A chip system includes a processor for calling and running a computer program from a memory, so that a communication device equipped with the chip system executes a method for identifying factors affecting the acceptance capacity of a power grid under multi-modal loads.

[0078] The present invention has the following beneficial effects: The proposed method for identifying factors influencing the grid's acceptance capacity under multi-modal load access systematically considers the multifaceted impacts of different load types on the grid. By comprehensively analyzing multiple factors, such as load fluctuations, power factor, voltage stability, and frequency fluctuations, and integrating them with timing characteristics, the method identifies the factors influencing each load's acceptance capacity. This method provides a comprehensive and accurate basis for evaluating grid load access, effectively optimizes load scheduling and planning, and enhances the grid's adaptive and intelligent regulation capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 Schematic diagram of the overall method of the present invention. DETAILED DESCRIPTION

[0080] The following is a combination of the embodiments of the present invention Figure 1 , the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the embodiments described are only part of the embodiments of the present invention, rather than all the embodiments. Unless otherwise specified, the technical means used in the embodiments are conventional means well known to those skilled in the art.

[0081] The present application provides a method and device for identifying factors affecting the acceptance capacity of a power grid under multi-form loads. The method for identifying factors affecting the acceptance capacity of a power grid under multi-form load access proposed by the present invention comprehensively analyzes the impact of various loads on the power grid, combines improved path analysis methods, multi-dimensional extended space learning and other methods to process the typical characteristics of various loads, and predicts the operating performance under multiple load access, providing a new technical path for improving the acceptance capacity of the power grid, which can effectively solve the shortcomings of the existing technology and improve the load access capacity and operating efficiency of the power grid. First, multi-scale data such as industrial electricity consumption, agricultural electricity consumption, transportation electricity consumption, municipal life electricity consumption and new energy power generation are obtained in the form of years, months and days. Secondly, feature extraction and cluster analysis are performed on the obtained power load data through improved path integral method and multi-dimensional extended space method. Then, the correlation between the acceptance capacity of the power grid and the clustered feature data is analyzed by improved Spearson method. Finally, the improved LSTM algorithm is used to predict the operating performance under multiple load access scenarios, so as to achieve the identification of factors affecting the acceptance capacity of the power grid under multi-form load access.

[0082] like Figure 1 As shown in FIG, the method for identifying factors affecting the grid acceptance capacity under multi-form loads includes the following steps:

[0083] Step 1: Obtain various types of multi-scale load data;

[0084] As the supply system that supports human life and economic and social development, the power system is characterized by widespread geographical distribution, real-time balance between power generation and consumption, massive amounts of energy transmission, power transmission at the speed of light, highly reliable communications and dispatching, continuous real-time operation, and the instantaneous expansion of major accidents. Furthermore, the intermittent and random nature of renewable energy sources can lead to drastic fluctuations in power flows, reducing the predictability of grid dispatch. These issues determine that the data generated by power system operation has the characteristics of big data. Through power distribution and utilization management systems, user information collection systems, marketing service systems, and geographic information systems, we can obtain multi-scale data on industrial electricity consumption, agricultural electricity consumption, transportation electricity consumption, municipal electricity consumption, and renewable energy generation at annual, monthly, and daily scales.

[0085] Step 2: Extract various types of multi-scale load data features using the improved path multi-integral method;

[0086] Connecting the various types of multi-scale load data obtained in step 1 in chronological order can be regarded as a special case of the path. Reconstruct the data and expand it to the r-dimensional path space to obtain a continuous mapping X. The coordinates of the r-dimensional path at time t are obtained as follows:

[0087]

[0088] in: is the coordinate of the r-dimensional path at time t, r is the dimension, which is a natural number, and t is the time series of the data.

[0089] Convert the superscripts of the mapping X into a sequence

[0090] i1,i2,...,i k ∈{1,...,r}

[0091] Where: i k is the kth type of load data, k is the total number of load types, and its value is {k|k∈Z, 1≤k≤4}.

[0092] Then the one-dimensional k-fold integral is expressed as:

[0093]

[0094] Where: T is the total time, F() is the integral function of each type of multi-scale load data, and d is the differential symbol.

[0095] For different types of loads, the r-dimensional multi-integral is expressed as:

[0096]

[0097] After solving and sorting, the characteristic dimensions of various loads are obtained as follows:

[0098]

[0099] Since the truncated path multi-integral is used as the feature introduction, the zeroth term in the above formula is a constant and can be removed. Therefore, its feature dimension is reduced to:

[0100]

[0101] in: is the multi-scale load data feature; r is the dimension of the data, such as converting one-dimensional data into high-dimensional data features such as three-dimensional space vectors; k is the type of load data, including industrial load, renewable energy load, electric vehicle charging load, etc.

[0102] Step 3: Cluster analysis of the features in step 2 based on the multidimensional extended space learning method

[0103] For the feature data obtained in step 2 The feature dimension is r. Since there is real-time data in the sample data and the total sample size is large, in order to obtain the clustering of new data in real time, the projection relationship between historical data and new data is established through the learning method of the extended space. First, set To add new data, and P k Projected to In the r-dimensional subspace formed, the cluster labels are obtained, and the objective function is:

[0104]

[0105] in: For the i-th newly added series of the k-th load type, is the multi-scale load data feature obtained in step 2; is the optimal projection vector of the kth load, In order to select a variety of matrix norms to constrain the self-representation matrix, γ1 and γ2 are trade-off parameters.

[0106] By minimizing the norm, the projection between historical data and new data is established. The optimal projection vector is:

[0107]

[0108] Where: E is a unit vector.

[0109] After obtaining the optimal projection, it is still necessary to obtain the cluster labels of the newly added data, so the sample and The residual of Zhang Cheng's subspace in the jth dimension is:

[0110]

[0111] in: For samples and The residual of the spanned subspace in the jth dimension; is the jth non-zero element in the optimal vector.

[0112] According to the above residuals, by comparing the samples The residual value in each dimension determines the minimum residual dimension as the cluster label:

[0113]

[0114] in: is the cluster label of the data, and arg(*) is the complex argument principal value function.

[0115] Get the clustered data for:

[0116]

[0117] Step 4: Use the improved Spearson method to analyze the correlation between the grid acceptance capacity and the clustered characteristic data;

[0118] Through the above steps to analyze the factors affecting load, some of the influencing factors have been preliminarily determined. However, the correlation between the factors such as load volatility, the impact of load on voltage stability during peak periods and the grid's acceptance capacity still needs to be quantitatively analyzed to determine.

[0119] The feature data obtained in step 3 The grid acceptance capacity data obtained from step 1 Find the mean of each:

[0120]

[0121] in: is the mean value of the influencing factor data, is the mean value of the grid acceptance capacity data;

[0122] Find the population covariance based on the mean:

[0123]

[0124] in: is the covariance of influencing factors and grid acceptance capacity data;

[0125] Calculate the correlation coefficient of the two types of data respectively:

[0126]

[0127] in: and are the variances of influencing factors and grid acceptance capacity data respectively;

[0128] The Pearson correlation coefficient is obtained by calculation:

[0129]

[0130] in: is the Pearson correlation coefficient between the influencing factors and the grid acceptance capacity data.

[0131] Since both load data and grid acceptance capacity data are nonlinear, the clustered multi-scale load data is converted into hierarchical data. The Spearman coefficient is obtained by calculating the correlation coefficient of the rank data:

[0132]

[0133] in: is the Spearman correlation coefficient between the influencing factors and the grid acceptance capacity data.

[0134] Combining the Pearson coefficient and the Spearman coefficient, we get the improved Spearman correlation coefficient:

[0135]

[0136] in: is the multi-scale load data feature after clustering, is the data of the maximum impact that the power grid can withstand, α1 is the proportional coefficient of Pearson correlation, α2 is the proportional coefficient of improved Spearman correlation, and α1+α2=1, which is determined according to the actual project; is the correlation coefficient obtained by the Pearson correlation method, is the correlation coefficient obtained by the Spearman correlation method.

[0137] Step 5: Use the improved LSTM algorithm to predict the grid's operating performance under various load access scenarios;

[0138] Based on historical data The feature data obtained in step 3 Conduct static graph learning and dynamic graph learning separately:

[0139]

[0140] in: A H ,A G are learnable parameters, H and G are query parameters and path parameters in the attention mechanism respectively, and L s is the adjacency matrix of the static graph. The ReLU function is used to eliminate negative connections in the matrix. The Softmax function normalizes each row of the matrix to preserve the relative relationship of the data:

[0141]

[0142] Where: ∥*∥ is the Euclidean norm; DistNet(*) is the distance network; L d is the adjacency matrix of the dynamic graph.

[0143] Through convolution operations, the clustered multi-scale load data features are combined with the multi-scale load data features learned through static and dynamic graphs to improve the diffusion capacity of node information on static and dynamic graphs and prevent nodes from losing their own unique information during the information propagation process:

[0144]

[0145] in: and is the data of the feature sequence after static graphics and dynamic graphics learning; β1 and β2 are hyperparameters; W1 and W2 are learnable parameters, and b1 and b2 are learning biases.

[0146] The prediction of the operating status of the power grid is:

[0147]

[0148] Where: W′1 and W′2 are learnable parameters of the fully connected layer, b′1 and b′2 are learnable biases, is the predicted value of the power grid operation status.

[0149] The above steps are integrated into the device for identifying factors affecting the grid's acceptance capacity under multi-modal loads. The device includes four modules: feature extraction module, feature clustering module, correlation analysis module, and prediction module. The following is the work content of each module:

[0150] Feature extraction module: First, it communicates with the power grid system to obtain various types of multi-scale load data, extracts data features through the improved path multi-integration method, and finally passes it to the feature clustering analysis module.

[0151] Feature clustering module: Use the multidimensional extended space learning method to cluster the typical features in step 1 and pass the results to the correlation analysis module.

[0152] Correlation Analysis Module: This module conducts time series analysis on grid and load data, evaluates the correlation between the temporal characteristics of load changes and the grid's acceptance capacity, and conducts a comprehensive analysis based on various indicators of grid acceptance capacity (such as grid operating costs, voltage stability, load carrying capacity, etc.) to optimize the grid's acceptance strategy.

[0153] Prediction module: By analyzing historical load data, it predicts the future load demand change trend of different types of loads (such as industrial load, residential load, electric vehicle charging load, renewable energy load, etc.).

[0154] A computer-readable storage medium stores a computer program. When the computer program is run on a computer, the computer is enabled to execute a method for identifying factors affecting the acceptance capacity of a power grid under multi-modal loads.

[0155] A chip system includes a processor for calling and running a computer program from a memory, so that a communication device equipped with the chip system executes a method for identifying factors affecting the acceptance capacity of a power grid under multi-modal loads.

[0156] The proposed method for identifying factors influencing the grid's acceptance capacity under multi-modal load access systematically considers the multifaceted impact of different load types on the grid. By comprehensively analyzing factors such as load fluctuation, power factor, voltage stability, and frequency fluctuation, and integrating them with timing characteristics, the method identifies the factors influencing each load's acceptance capacity. This method provides a comprehensive and accurate basis for evaluating grid load access, effectively optimizes load scheduling and planning, and enhances the grid's adaptive and intelligent regulation capabilities.

[0157] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various deformations, modifications, and substitutions made to the technical solutions of the present invention by ordinary technicians in this field should fall within the scope of protection determined by the claims of the present invention.

Claims

1. A method for identifying factors affecting the grid's acceptance capacity under multi-form loads, characterized in that: include: Step 1: Obtain various types of multi-scale load data; Step 2: Extract various types of multi-scale load data features by the path multi-integral method; Step 3: Cluster analysis of multi-scale load data features based on multi-dimensional extended space learning method; Step 4: Use the improved Spearson method to analyze the correlation between the grid acceptance capacity and the characteristics of the clustered multi-scale load data; Step 5: Use the improved LSTM algorithm to predict the grid's operating performance under various load access scenarios; Wherein, step 2 includes: The various types of multi-scale load data obtained in step 1 are connected in chronological order as a special case of the path. The data are reconstructed and expanded to the r-dimensional path space to obtain a continuous mapping X. The coordinates of the r-dimensional path at time t are obtained as follows: in: is the coordinate of the r-dimensional path at time t, r is the dimension, and t is the time series of the data; Convert the superscripts of the mapping X into a sequence i1,i2,...,i k ∈{1,...,r} Where: i k is the kth load data, k is the total number of load types; The one-dimensional k-fold integral is expressed as: Where: T is the total time, F() is the integral function of each type of multi-scale load data, and d is the differential symbol; For different types of loads, the r-dimensional multi-integral is expressed as: The solution is: in: is the multi-scale load data feature; r is the dimension of the data, such as converting one-dimensional data into high-dimensional data features such as three-dimensional space vectors; k is the type of load data, including industrial load, renewable energy load, electric vehicle charging load, etc.

2. The method for identifying factors affecting the grid acceptance capacity under multi-form loads according to claim 1, characterized in that: Step 3 includes: For multi-scale load data characteristics set up To add new data, and P k Projected to In the r-dimensional subspace formed, the cluster labels are obtained, and the objective function is: in: For the i-th newly added series of the k-th load type, is the multi-scale load data feature; is the optimal projection vector of the kth load, It constrains the self-representation matrix for various matrix norms, and γ1 and γ2 are trade-off parameters; By minimizing the norm, the projection between historical data and new data is established. The optimal projection vector is: Where: E is a unit vector; Obtain and The residual of Zhang Cheng's subspace in the jth dimension is: in: for and The residual of the spanned subspace in the jth dimension; is the jth non-zero element in the optimal vector; According to the above residuals, by comparing The residual value in each dimension determines the minimum residual dimension as the cluster label: in: is the cluster label of the data, arg(*) is the complex argument principal value function; Get the multi-scale load data characteristics after clustering for:

3. The method for identifying factors affecting the grid acceptance capacity under multi-form loads according to claim 2, characterized in that: Step 4 includes: The multi-scale load data features after clustering and grid capacity data Find the mean of each: in: is the mean value of the influencing factor data, is the mean value of the grid acceptance capacity data; Find the population covariance based on the mean: in: is the covariance of influencing factors and grid acceptance capacity data; Calculate the correlation coefficient of the two types of data respectively: in: and are the variances of influencing factors and grid acceptance capacity data respectively; The Pearson correlation coefficient is obtained by calculation: in: is the Pearson correlation coefficient between the influencing factors and the grid acceptance capacity data; Convert the clustered multi-scale load data into hierarchical data The Spearman coefficient is obtained by calculating the correlation coefficient of the rank data: in: is the Spearman correlation coefficient between the influencing factors and the grid acceptance capacity data; By using the Pearson coefficient and the Spearman coefficient, the improved Spearman correlation coefficient is obtained: in: is the multi-scale load data feature after clustering, is the data of the maximum impact that the power grid can withstand, α1 is the proportional coefficient of Pearson correlation, and α2 is the proportional coefficient of improved Spearman correlation; is the correlation coefficient obtained by the Pearson correlation method, is the correlation coefficient obtained by the Spearman correlation method.

4. The method for identifying factors affecting the grid acceptance capacity under multi-form loads according to claim 3, characterized in that: Step 5 includes: Multi-scale load data features Conduct static graph learning and dynamic graph learning: in: A H ,A G are learnable parameters, H and G are query parameters and path parameters in the attention mechanism respectively, and L s is the adjacency matrix of the static graph. The ReLU function is used to eliminate negative connections in the matrix. The Softmax function normalizes each row of the matrix to preserve the relative relationship of the data: Where: ||*|| is the Euclidean norm; DistNet(*) is the distance network; L d is the adjacency matrix of the dynamic graph; Through convolution operations, the clustered multi-scale load data features are combined with the multi-scale load data features learned through static and dynamic graphs: in: and is the data of the feature sequence after static graph and dynamic graph learning; β1 and β2 are hyperparameters; W1 and W2 are learnable parameters, and b1 and b2 are learning biases; The prediction of the operating status of the power grid is: Where: W′1 and W′2 are learnable parameters of the fully connected layer, b′1 and b′2 are learnable biases, is the predicted value of the power grid operation status.

5. A device for identifying factors affecting the acceptance capacity of a power grid under multi-form loads, comprising: a method for identifying factors affecting the acceptance capacity of a power grid under multi-form loads according to any one of claims 1 to 4, characterized in that: include: Feature extraction module, feature clustering module, correlation analysis module and prediction module; among them: The feature extraction module is connected to the power grid to obtain various types of multi-scale load data, and extracts the features of various types of multi-scale load data through the path multi-integral method; The feature clustering module performs cluster analysis on multi-scale load data features based on the multi-dimensional extended space learning method; The correlation analysis module uses the improved Spearson method to analyze the correlation between the grid acceptance capacity and the multi-scale load data characteristics after clustering; The prediction module uses an improved LSTM algorithm to predict the operating performance of the power grid under various load access scenarios.

6. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is run on a computer, the computer is caused to execute the method according to any one of claims 1 to 4.

7. A chip system, characterized in that: It includes a processor for calling and running a computer program from a memory, so that a communication device equipped with the chip system executes the method according to any one of claims 1 to 4.

Images