Multi-region power load forecasting method, device, equipment, medium and program product

Abstract

The present invention discloses a multi-region power load forecasting method, device, equipment, medium and program product. The method includes: collecting historical load time series data of multiple regions in a power system, preprocessing the historical load time series data to obtain candidate load time series data, performing deep feature extraction and multidimensional feature analysis on the candidate load time series data, fusing the deep features with the multidimensional feature matrix to obtain target load features, and performing cluster analysis on multiple regions based on the target load features to determine regional clusters corresponding to each region, constructing load forecasting models corresponding to each regional cluster according to the target load features, and forecasting power load for the power system based on the load forecasting models corresponding to each regional cluster. The present invention effectively mines the load characteristics of each region, thereby better adapting to load changes in different regions and providing accurate power load forecasting for different regions in the power system.

Description

Technical Field

[0001] The present invention relates to the field of power load forecasting, and in particular to a multi-region power load forecasting method, device, equipment, medium and program product. Background Art

[0002] In recent years, the widespread integration of distributed energy resources (DREs) has provided users with clean, green electricity. However, as their penetration continues to increase, traditional distribution networks are struggling to accommodate this DRE. Frequent problems such as over-limit flow, power reverse flow, and protection misoperation have severely impacted the stability of distribution networks. Simultaneously, profound changes have also taken place on the demand side of electricity consumption. Loads are no longer simply passively subject to one-way control by the grid, but are increasingly capable of bidirectional interaction with the grid, resulting in a more diverse range of control methods. Furthermore, the development and coordinated operation of series-parallel combined flexible energy storage devices on the distribution side are also evolving, placing higher demands on grid flexibility. Consequently, traditional distribution networks are unable to adapt to and meet the complex and diverse power control requirements of future grids. Furthermore, in distributed energy storage systems, power and energy demands, as well as user consumption behaviors, vary across regions, creating challenges for optimizing resource scheduling. Therefore, ensuring secure and stable power supply and rational resource scheduling of DREs requires reasonable forecasting of regional power loads and determining energy demand. Current distributed load forecasting methods often use a single algorithm to forecast power load for different regions, thus ignoring the uniqueness and differences of load data in different regions of the power system, resulting in low power load forecasting accuracy for the power system, which seriously affects the efficiency of power resource scheduling. Summary of the Invention

[0003] The main purpose of the present invention is to provide a multi-region power load forecasting method, device, equipment, medium and program product, aiming to solve the technical problem that the existing technology ignores the uniqueness and differences of load data in different regions of the power system, resulting in low power load forecasting accuracy of the power system and seriously affecting the efficiency of power resource scheduling.

[0004] To achieve the above object, the present invention provides a multi-region power load forecasting method, the method comprising the following steps:

[0005] Collecting historical load time series data of multiple areas in the power system, and preprocessing the historical load time series data to obtain candidate load time series data;

[0006] Performing deep feature extraction on the candidate load time series data to obtain a deep feature matrix for each region;

[0007] Performing multi-dimensional feature analysis on the candidate load time series data to obtain a multi-dimensional feature matrix for each region;

[0008] Performing feature fusion on the deep feature matrix and the multidimensional feature matrix to obtain target load features, and performing cluster analysis on the multiple regions based on the target load features to determine a region cluster corresponding to each region;

[0009] Constructing a load forecasting model corresponding to each regional cluster according to the target load characteristics;

[0010] The power load of the power system is predicted based on the load prediction model corresponding to each regional cluster.

[0011] Optionally, performing deep feature extraction on the candidate load time series data to obtain a deep feature matrix of each region includes:

[0012] Performing a bidirectional long-term and short-term dependency analysis on the candidate load time series data to obtain a forward hidden state and a reverse hidden state of each time step, and concatenating the forward hidden state with the reverse hidden state to obtain a bidirectional dependency feature of each time step;

[0013] Performing linear transformation based on the bidirectional hidden state to construct a query matrix, a key matrix, and a value matrix;

[0014] Determine the attention features for each time step based on the query matrix, the key matrix, and the value matrix:

[0015] ;

[0016] in, Represents the time step The attention characteristics of represents the query matrix, represents the matrix, represents the value matrix, represents transpose, represents the dimension of the bond matrix, represents the activation function;

[0017] Concatenate the bidirectional dependency features and attention features of each time step to obtain the initial deep features:

[0018] ;

[0019] in, Represents the time step The initial depth features of represents the feature concatenation operation, Represents the time step Bidirectional dependency characteristics;

[0020] Construct a deep feature matrix based on the initial deep features of each time step:

[0021] ;

[0022] in, represents the deep feature matrix, represents the time dimension, Represents the feature dimension.

[0023] Optionally, the candidate load time series data is subjected to a multi-dimensional feature analysis to obtain a multi-dimensional feature matrix of each region:

[0024] Calculate the mean, standard deviation, and coefficient of variation of the candidate load time series data:

[0025] ;

[0026] in, represents the mean of candidate load time series data, represents the standard deviation of the candidate load time series data, represents the coefficient of variation;

[0027] Performing global load fluctuation analysis based on the mean, the standard deviation, and the coefficient of variation to obtain basic statistical characteristics;

[0028] Perform time fluctuation analysis based on the candidate load time series data to obtain time pattern characteristics, which include peak-to-valley difference, time period average ratio, and daily load rate:

[0029] ;

[0030] ;

[0031] ;

[0032] in, Indicates the peak-to-valley difference of load within a day. Indicates the maximum load in a day. Indicates the minimum load value in a day. It represents the load average ratio between morning and evening hours. represents the average load value during the morning period, Indicates the average load value during the night time. represents the daily load rate, Indicates the average load value within a day;

[0033] A short-term fluctuation analysis is performed based on the candidate load time series data to obtain a volatility feature, wherein the volatility feature includes a lagged autocorrelation coefficient:

[0034] ;

[0035] in, represents the lagged autocorrelation coefficient, Indicates the The load value of a time step is a data point in the candidate load time series data. represents the total number of time steps, represents the number of lag steps, represents the data mean of the candidate load time series data;

[0036] The basic statistical features, the time pattern features and the volatility features are combined into a multi-dimensional feature matrix.

[0037] Optionally, the performing feature fusion on the deep feature matrix and the multidimensional feature matrix to obtain target load features, and performing cluster analysis on the multiple regions based on the target load features to determine region clusters corresponding to the respective regions, includes:

[0038] The deep feature matrix and the multi-dimensional feature matrix are subjected to feature fusion to obtain the target load feature:

[0039] ;

[0040] in, represents the target load characteristics, Represents a multidimensional feature matrix;

[0041] Determine the feature similarity between regions based on the target load feature:

[0042] ;

[0043] in, Indicates area and region The feature similarity between The characteristic dimension representing the target load characteristics, Indicates area No. Features, Indicates area No. Features

[0044] Cluster analysis is performed on the multiple regions according to the feature similarity to determine a region cluster corresponding to each region.

[0045] Optionally, constructing a load forecasting model corresponding to each regional cluster according to the target load characteristics includes:

[0046] Construct an initial prediction model corresponding to each regional cluster. The objective function of the initial prediction model is as follows:

[0047] ;

[0048] in, represents the objective function, represents the loss function, represents the number of samples, Indicates the The true value of the sample, Indicates the Samples passed The predicted value after the decision tree, represents a set of decision trees, represents the regularization term used to prevent overfitting, Indicates the The complexity function of a decision tree;

[0049] The target load characteristics are normalized:

[0050] ;

[0051] in, Represents the normalized The first sample Features, represents the first The first sample Features, and Respectively represent The maximum and minimum values ​​of the features;

[0052] The importance of each feature in the normalized target load feature is evaluated, and the normalized target load feature is screened based on the importance evaluation result to obtain the target training feature. The importance evaluation refers to the following formula:

[0053] ;

[0054] in, Represents a decision tree The gain value of the feature in , Indicates the result of materiality assessment;

[0055] The initial prediction model is trained based on the target training features to obtain a load prediction model.

[0056] Optionally, preprocessing the historical load time series data to obtain candidate load time series data includes:

[0057] Calculate the standard score for each data point in the historical load time series data:

[0058] ;

[0059] in, Represents a data point The standard score of represents the mean of the historical load time series data, Represents the standard deviation of historical load time series data;

[0060] Marking data points with standard scores higher than an abnormal threshold in the historical load time series data as abnormal data points, and performing data cleaning on the abnormal data points;

[0061] Fill missing data in the historical load time series data after data cleaning, obtain the initial load time series data, and fill missing data according to the following formula:

[0062] ;

[0063] in, Indicates the length of the data missing segment, Indicates the missing position of the current missing value in the missing segment, represents the known data points after the missing segment, represents the known data points before the missing segment;

[0064] The initial load time series data is normalized to obtain candidate load time series data. The normalization process is performed according to the following formula:

[0065] ;

[0066] in, represents the normalized data points, represents the minimum data point in the initial load time series data, Indicates the maximum value data point.

[0067] In addition, to achieve the above-mentioned object, the present invention further proposes a multi-region power load forecasting device, the multi-region power load forecasting device comprising:

[0068] A data processing module is used to collect historical load time series data of multiple areas in the power system and preprocess the historical load time series data to obtain candidate load time series data;

[0069] A deep feature analysis module is used to extract deep features from the candidate load time series data to obtain a deep feature matrix for each region;

[0070] A multidimensional feature analysis module is used to perform multidimensional feature analysis on the candidate load time series data to obtain a multidimensional feature matrix for each region;

[0071] a feature clustering analysis module, configured to perform feature fusion on the deep feature matrix and the multidimensional feature matrix to obtain target load features, and to perform cluster analysis on the multiple regions based on the target load features to determine a region cluster corresponding to each region;

[0072] A prediction model building module, configured to build a load prediction model corresponding to each regional cluster according to the target load characteristics;

[0073] The power load forecasting module is used to forecast the power load of the power system based on the load forecasting model corresponding to each regional cluster.

[0074] In addition, to achieve the above-mentioned purpose, the present application also proposes a multi-region power load forecasting device, which includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the computer program is configured to implement the steps of the multi-region power load forecasting method as described above.

[0075] In addition, to achieve the above-mentioned purpose, the present application also proposes a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the multi-region power load forecasting method as described above are implemented.

[0076] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the multi-region power load forecasting method as described above.

[0077] The present invention collects historical load time series data of multiple areas in the power system, preprocesses the historical load time series data, obtains candidate load time series data, performs deep feature extraction on the candidate load time series data, obtains a deep feature matrix of each area, performs multidimensional feature analysis on the candidate load time series data, obtains a multidimensional feature matrix of each area, fuses the deep feature matrix with the multidimensional feature matrix, obtains target load features, and performs cluster analysis on the multiple areas based on the target load features to determine the regional clusters corresponding to each area, constructs a load forecasting model corresponding to each regional cluster according to the target load features, and forecasts the power system based on the load forecasting model corresponding to each regional cluster. The system performs power load forecasting; since the present invention performs deep feature analysis and multi-dimensional feature analysis on the load time series data of each region respectively, it can realize in-depth mining of the load characteristics of each region, effectively avoiding the problem of ignoring the potential characteristics and differences of different regions, and improving the prediction accuracy. By fusing the deep features and multi-dimensional features, clustering analysis is performed based on the target load features obtained by the fusion, clustering of the same type of regions is achieved, and a reasonable load forecasting model is provided for the regions with the same characteristics, effectively improving the power load forecasting efficiency, more accurately capturing and reflecting the load change trend of each region, greatly improving the accuracy of the power system load forecast, and providing accurate forecast result support for power resource scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0079] Figure 1 It is a schematic structural diagram of a multi-region power load forecasting device in a hardware operating environment involved in an embodiment of the present invention;

[0080] Figure 2 This is a flow chart of a first embodiment of a multi-region power load forecasting method according to the present invention;

[0081] Figure 3 Schematic diagram of the deep feature extraction process in an embodiment of the multi-region power load forecasting method of the present invention;

[0082] Figure 4 A schematic diagram of a multi-dimensional feature analysis process in an embodiment of a multi-region power load forecasting method of the present invention;

[0083] Figure 5 A schematic diagram of a load forecasting model construction process in an embodiment of a multi-region power load forecasting method of the present invention;

[0084] Figure 6 This is a structural block diagram of the first embodiment of the multi-region power load forecasting device of the present invention.

[0085] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0086] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0087] Reference Figure 1 , Figure 1 This is a schematic diagram of the structure of a multi-region power load forecasting device in the hardware operating environment involved in an embodiment of the present invention.

[0088] like Figure 1 As shown, the multi-region power load forecasting device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display and an input unit, such as a keyboard. Optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a wireless fidelity (WI-FI) interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also be a storage device independent of the processor 1001.

[0089] Those skilled in the art will understand that Figure 1 The structure shown in the figure does not constitute a limitation on the multi-region power load forecasting device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0090] like Figure 1 As shown, the memory 1005 as a computer-readable storage medium may include an operating system, a network communication module, a user interface module, and a multi-region power load prediction program.

[0091] exist Figure 1In the multi-region power load forecasting device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the multi-region power load forecasting device of the present invention can be set in the multi-region power load forecasting device, and the multi-region power load forecasting device calls the multi-region power load forecasting program stored in the memory 1005 through the processor 1001, and executes the multi-region power load forecasting method provided by the embodiment of the present invention.

[0092] The embodiment of the present invention provides a multi-region power load forecasting method, referring to Figure 2 , Figure 2 FIG. 1 is a flow chart of a first embodiment of a multi-region power load forecasting method according to the present invention.

[0093] In this embodiment, the multi-region power load forecasting method includes the following steps:

[0094] Step S10: collecting historical load time series data of multiple areas in the power system, and preprocessing the historical load time series data to obtain candidate load time series data.

[0095] It should be noted that this embodiment can be applied to the power load forecasting of a distributed power supply system including multiple power consumption areas. This embodiment performs deep feature analysis and multi-dimensional feature analysis on the load time series data of each area respectively, thereby realizing in-depth mining of the load characteristics of each area, effectively avoiding the problem of ignoring the potential characteristics and differences of different areas, and improving the prediction accuracy. By fusing the deep features and multi-dimensional features, clustering analysis is performed based on the target load characteristics obtained by fusion, and clustering of the same type of areas is achieved. A reasonable load forecasting model is provided for areas with the same characteristics, effectively improving the power load forecasting efficiency, more accurately capturing and reflecting the load change trends of each area, greatly improving the accuracy of the power system load forecast, and providing accurate forecasting result support for power resource scheduling.

[0096] It should be understood that the execution subject of this embodiment may be a computing service device with data processing, network communication, and program execution capabilities, such as a tablet computer, personal computer, or mobile phone, or a terminal electronic device capable of performing the aforementioned functions. This embodiment and the following embodiments will be described below using a multi-region power load forecasting device (hereinafter referred to as the forecasting device) as an example.

[0097] It should be noted that a power system can be a distributed, multi-regional system. A multi-regional system emphasizes the system's coverage of multiple geographic or functional areas, each with independent power demand and supply management. This system can include central power stations or distributed generation units, delivering electricity to each consumption area via the grid.

[0098] It should be noted that this embodiment can perform short-term load forecasting on the power system. Short-term load forecasting of the power system refers to the prediction of changes in load demand of the power system in the next few hours, a day, a week or less. It is an important link in the operation and dispatching of the power system, and is of great significance for ensuring the safe and stable operation of the power grid, optimizing the allocation of power resources, and improving the quality of power supply services. Short-term load forecasting of the power system needs to take into account a variety of factors, including historical load data, weather changes, holidays, economic activities, social events, etc. Through comprehensive analysis of these factors, a mathematical model can be established or a machine learning algorithm can be used to predict future load demand. The results of short-term load forecasting can provide an important basis for power system dispatching, helping dispatchers to reasonably arrange power generation plans and adjust the operation mode of the power grid to cope with changes in load.

[0099] It is understood that in this embodiment, in the first step of load forecasting, the forecasting device can collect historical load time series data for various regions. This data is the basis for subsequent analysis, so the integrity and accuracy of the data must be ensured. After collection, the historical load time series data is preprocessed, for example, by cleaning the data and filling in missing data.

[0100] In some embodiments, when the prediction device collects data, it resamples the non-uniformly sampled data to a fixed interval (e.g., 15 minutes):

[0101] ;

[0102] in, Represents a set of time steps within a time window, used to describe the data distribution within a specific time window. Indicates the current time point, Indicates the half-width of the time window.

[0103] Furthermore, in order to improve data quality and thus improve prediction efficiency and accuracy, the above step S10 may include:

[0104] Step S101: Calculating the standard score of each data point in the historical load time series data;

[0105] Step S102: marking data points with standard scores higher than an abnormal threshold in the historical load time series data as abnormal data points, and performing data cleaning on the abnormal data points;

[0106] Step S103: Fill missing data in the historical load time series data after data cleaning to obtain initial load time series data;

[0107] Step S104: normalizing the initial load time series data to obtain candidate load time series data.

[0108] It's important to note that after collecting historical load time series data, the forecasting equipment can perform data cleaning to address outliers, such as sudden increases or decreases, or values ​​that clearly contradict common sense, to ensure data reliability. For missing data, the forecasting equipment can use appropriate methods, such as interpolation or extrapolation based on historical data, to supplement it. It also ensures that all data points are aligned for subsequent analysis. Finally, to eliminate dimensional differences between the data, the forecasting equipment can perform data standardization to ensure comparability and consistency.

[0109] It should be noted that the Z-score method can be used to clean the outliers in the data. The data points are marked as outliers and cleaned up. The standard score is calculated according to the following formula:

[0110] ;

[0111] in, Represents a data point The standard score of represents the mean of the historical load time series data, represents the standard deviation of the historical load time series data, is the abnormal threshold.

[0112] When missing values ​​occur, linear interpolation can be used to fill them. Refer to the following formula to fill missing data:

[0113] ;

[0114] in, Indicates the length of the data missing segment, Indicates the missing position of the current missing value in the missing segment, represents the known data points after the missing segment, Represents the known data point preceding the missing segment.

[0115] When the prediction device performs data scaling, the min-max normalization method can be used to scale the value to [0, 1]. The normalization process refers to the following formula:

[0116] ;

[0117] in, represents the normalized data points, represents the minimum data point in the initial load time series data, Indicates the maximum value data point.

[0118] After the above processing, normalized, aligned and complete time series data can be obtained. Indicates the The preprocessed load value at each time point:

[0119] ;

[0120] Step S20: performing deep feature extraction on the candidate load time series data to obtain a deep feature matrix of each region.

[0121] In some embodiments, the forecasting device can construct a deep feature extraction network to perform deep feature extraction on the candidate load time series data, obtaining a deep feature matrix for each region. The deep feature extraction network can be constructed based on a bidirectional long short-term memory (BiLSTM) network and an attention mechanism module. The deep feature extraction network can automatically learn the high-order temporal dependencies and nonlinear patterns of the load data, overcoming the limitations of traditional feature engineering. We use a multi-layer recurrent neural network architecture, combined with a bidirectional long short-term memory network and an attention mechanism, to fully exploit the deep features of the load data.

[0122] Furthermore, in order to accurately extract the deep features in the load data and explore the potential load characteristics of each area, refer to Figure 3 , Figure 3 This is a schematic diagram of a deep feature extraction process in one embodiment. Step S20 may include:

[0123] Step S201: performing a bidirectional long-term and short-term dependency analysis on the candidate load time series data, obtaining a forward hidden state and a reverse hidden state of each time step, and concatenating the forward hidden state and the reverse hidden state to obtain a bidirectional dependency feature of each time step;

[0124] Step S202: performing linear transformation based on the bidirectional hidden state to construct a query matrix, a key matrix, and a value matrix;

[0125] Step S203: determining the attention feature of each time step according to the query matrix, the key matrix and the value matrix;

[0126] Step S204: concatenate the bidirectional dependency features and attention features of each time step to obtain the initial depth features;

[0127] Step S205: constructing a depth feature matrix based on the initial depth features of each time step.

[0128] In some embodiments, the prediction device may receive the pre-processed candidate load time series data through the input layer of the deep feature extraction network, and may integrate category features (such as region, season, etc.) as auxiliary inputs, and learn long-term and short-term dependencies through a bidirectional long short-term memory network, referring to the following formula:

[0129] ;

[0130] The prediction device assigns the weight of the load data at each time step through the self-attention mechanism module to enhance the weight of key time points, referring to the following formula:

[0131] ;

[0132] in, Represents the time step The attention characteristics of represents the query matrix, represents the matrix, represents the value matrix, represents transpose, represents the dimension of the bond matrix, Represents the activation function.

[0133] The prediction device concatenates the bidirectional dependency features and attention features of each time step:

[0134] ;

[0135] in, Represents the time step The initial depth features of represents the feature concatenation operation, Represents the time step The bidirectional dependency feature.

[0136] The prediction device constructs a depth feature matrix based on the initial depth features of each time step:

[0137] ;

[0138] in, represents the deep feature matrix, represents the time dimension, Represents the feature dimension.

[0139] In some embodiments, the prediction device may use a loss function that combines MAE and Huber Loss to train the deep feature extraction network to balance the robustness and accuracy of the deep feature extraction network:

[0140] ;

[0141] in, represents the loss value, represents the weight coefficient, represents the average error function, represents the smooth loss function, represents the predicted value, Represents the true value.

[0142] Dynamic feature selection of deep feature extraction networks can dynamically adjust input features based on gradient weight analysis (Grad-CAM) or feature importance scores.

[0143] Dropout and Layer Normalization can be used to regularize deep feature extraction networks to prevent overfitting.

[0144] Step S30: performing multi-dimensional feature analysis on the candidate load time series data to obtain a multi-dimensional feature matrix for each region.

[0145] It should be noted that feature engineering is a key step in load forecasting, which is directly related to the accuracy and generalization ability of the forecasting model. This embodiment performs multi-dimensional feature extraction and deep feature extraction on the candidate load time series data respectively.

[0146] In some embodiments, the forecasting device can perform multidimensional feature extraction from three aspects: basic statistical features, temporal pattern features, and volatility features. Basic statistical features include the mean, standard deviation, and coefficient of variation of the load, which can reflect the overall level and fluctuation of the load. Temporal pattern features focus on the intra-day, intra-week, and seasonal variations of the load, such as peak and valley periods, differences between weekdays and weekends, and seasonal load variations. Volatility features describe the volatility and periodicity of the load through multi-time scale volatility, autocorrelation characteristics, and periodic strength indicators.

[0147] Furthermore, in order to accurately extract the multidimensional features of the load data in each region, Figure 4 , Figure 4 FIG. 5 is a schematic diagram of a multi-dimensional feature analysis process in one embodiment. Step S30 may include:

[0148] Step S301: Calculate the mean, standard deviation and coefficient of variation of the candidate load time series data;

[0149] Step S302: performing global load fluctuation analysis based on the mean, the standard deviation, and the coefficient of variation to obtain basic statistical characteristics;

[0150] Step S303: performing time fluctuation analysis based on the candidate load time series data to obtain time pattern characteristics;

[0151] Step S304: performing short-term fluctuation analysis based on the candidate load time series data to obtain fluctuation characteristics;

[0152] Step S305: combining the basic statistical features, the time pattern features, and the volatility features into a multi-dimensional feature matrix.

[0153] It's important to note that multidimensional feature analysis is a key step in load forecasting, directly impacting the accuracy and generalization capabilities of the forecasting model. In addition to leveraging deep feature extraction networks, forecasting equipment can also extract features from three aspects: basic statistical features, temporal pattern features, and volatility features.

[0154] It should be noted that basic statistical characteristics may include the mean, standard deviation, and coefficient of variation of the load, which can reflect the overall level and fluctuation of the load and are calculated according to the following formula:

[0155] ;

[0156] in, represents the mean of candidate load time series data, represents the standard deviation of the candidate load time series data, represents the coefficient of variation.

[0157] Temporal pattern characteristics focus on load variations within the day, week, and season, such as peak and valley periods, differences between weekdays and weekends, and seasonal load variations. In some embodiments, the temporal pattern characteristics include peak-valley difference, time period average ratio, and daily load rate, calculated according to the following formula:

[0158] ;

[0159] ;

[0160] ;

[0161] in, Indicates the peak-to-valley difference of load within a day. Indicates the maximum load in a day. Indicates the minimum load value in a day. It represents the load average ratio between morning and evening hours. Indicates the average load value during the morning period (e.g. 7:00-9:00). Indicates the average load value during the night period (e.g. 23:00-5:00). represents the daily load rate, Indicates the average load value for a day.

[0162] The volatility characteristics describe the volatility and periodicity of the load through multi-time scale volatility, autocorrelation characteristics, and periodic intensity indicators. The autocorrelation characteristics can be obtained based on the lagged autocorrelation coefficient, which is calculated according to the following formula:

[0163] ;

[0164] in, represents the lagged autocorrelation coefficient, Indicates the The load value of a time step is a data point in the candidate load time series data. represents the total number of time steps, represents the number of lag steps, Indicates the data mean of the candidate load time series data.

[0165] Finally, the extracted basic statistical features, deep feature matrix, and original load data are concatenated to form the input data of the model. For example, all features can be combined into a feature matrix F, where each row corresponds to a time point or time window:

[0166] ;

[0167] Step S40: performing feature fusion on the deep feature matrix and the multidimensional feature matrix to obtain target load features, and performing cluster analysis on the multiple regions based on the target load features to determine region clusters corresponding to each region.

[0168] It is understandable that in order to classify regions with similar load characteristics, this embodiment can perform cluster analysis on multiple regions in the power system, for example, using a hierarchical clustering method with Ward's minimum variance method as the link criterion and similarity measurement based on Euclidean distance.

[0169] Furthermore, in order to accurately aggregate regions of the same type and thus improve prediction efficiency, the above step S40 may include:

[0170] Step S401: performing feature fusion on the deep feature matrix and the multi-dimensional feature matrix to obtain target load features;

[0171] Step S402: determining the feature similarity between the regions based on the target load feature;

[0172] Step S403: performing cluster analysis on the multiple regions according to the feature similarity to determine a region cluster corresponding to each region.

[0173] It should be noted that this embodiment calculates the feature similarity between regions based on the target load characteristics by fusing the deep feature matrix and the multidimensional feature matrix, thereby effectively identifying areas with similar load characteristics and providing a basis for the subsequent construction of a differentiated prediction model. The feature fusion is referred to the following formula:

[0174] ;

[0175] in, represents the target load characteristics, Represents a multidimensional feature matrix.

[0176] The feature similarity between regions is calculated according to the following formula:

[0177] ;

[0178] in, Indicates area and region The feature similarity between The characteristic dimension representing the target load characteristics, Indicates area No. Features, Indicates area No. Features.

[0179] In some embodiments, during the clustering process, the prediction device can dynamically adjust the number of clusters to ensure the rationality and accuracy of the clustering. The prediction device can also optimize feature weights to improve the stability and reliability of the clustering results. Finally, the quality of the clustering results is evaluated through stability verification to ensure the effectiveness and repeatability of the clustering.

[0180] In some embodiments, K data points are randomly selected as initial cluster centers, denoted as , for each sample , calculate its distance from each cluster center and assign it to the cluster corresponding to the nearest cluster center. , calculate the mean of all samples in the cluster , as the new cluster center:

[0181] ;

[0182] in, It is a cluster The number of samples in .

[0183] Step S50: constructing a load prediction model corresponding to each regional cluster according to the target load characteristics.

[0184] In some embodiments, the prediction device can build differentiated prediction models based on the load characteristics of each category. The model selection strategy is based on the selection of adaptive prediction algorithms based on category features, such as XGBoost, LightGBM, CatBoost, etc. These algorithms have advantages in processing complex data and capturing nonlinear relationships. In some embodiments, the prediction device can optimize the model parameters in a targeted manner according to the characteristics of each category area to improve the accuracy of the prediction. In terms of feature input, the category features are used as a supplement to the model input, and feature importance analysis is used to determine which features have the greatest impact on the prediction results. At the same time, the prediction device can adopt a dynamic feature selection mechanism to dynamically adjust the input features according to the performance of the model and changes in the data to ensure the stability and adaptability of the prediction model.

[0185] Furthermore, in order to improve the accuracy of load forecasting and accurately capture and reflect the load change trend of each region, Figure 5 , Figure 5 FIG. 5 is a schematic diagram of a load forecasting model construction process in an embodiment. Step S50 may include:

[0186] Step S501: constructing an initial prediction model corresponding to each regional cluster;

[0187] Step S502: normalizing the target load characteristics;

[0188] Step S503: performing importance evaluation on each feature in the normalized target load feature, and screening the normalized target load feature based on the importance evaluation result to obtain target training features;

[0189] Step S504: training the initial prediction model based on the target training features to obtain a load prediction model.

[0190] It is understandable that, in order to more accurately predict load changes in various regions, this embodiment employs a method of modeling and extracting features for each region separately. Because load data characteristics vary significantly across regions, a single prediction model is unlikely to comprehensively and accurately reflect the load variation patterns across all regions. Therefore, this embodiment constructs multiple load models based on the load data characteristics of each region. These models fully account for the unique characteristics of each region, thereby more accurately capturing and reflecting the load variation trends in each region.

[0191] In a specific implementation, the prediction device can prepare characteristic data, including historical power load data and characteristic data of target load characteristics: Power load data: ,in Indicates the i Load value at a time point; characteristic data: ,in It is i The feature vectors at each time point need to be normalized during data preprocessing to eliminate their dimensional differences:

[0192] ;

[0193] in, Represents the normalized The first sample Features, represents the first The first sample Features, and Respectively represent The maximum and minimum values ​​of the features.

[0194] The prediction device can build an initial prediction model based on XGBoost and use the feature importance evaluation function of XGBoost to screen the features that have the greatest impact on power load prediction. The importance evaluation refers to the following formula:

[0195] ;

[0196] in, Represents a decision tree The gain value of the feature in , Indicates the result of importance assessment.

[0197] XGBoost is an ensemble learning algorithm based on gradient boosting. It improves model learning efficiency by building multiple weak learners (usually decision trees) and combining them into a strong learner. Its goal is to minimize the loss function. The objective function of the initial prediction model is as follows:

[0198] ;

[0199] in, represents the objective function, represents the loss function, represents the number of samples, Indicates the The true value of the sample, Indicates the Samples passed The predicted value after the decision tree, represents a set of decision trees, represents the regularization term used to prevent overfitting, Indicates the The complexity function of a decision tree.

[0200] It is i The predicted value of a sample is obtained by summing the predicted values ​​of T trees:

[0201] ;

[0202] In each iteration, XGBoost optimizes the loss function by calculating the first and second derivatives:

[0203] ;

[0204] ;

[0205] For the squared error loss function:

[0206] ;

[0207] ;

[0208] Step S60: performing power load forecasting on the power system based on the load forecasting model corresponding to each regional cluster.

[0209] In some embodiments, the prediction device may use metrics such as root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) to evaluate model performance, as calculated by the following formula:

[0210] ;

[0211] ;

[0212] ;

[0213] in, Indicates the actual value, Represents the model prediction value. After calculating the prediction error of the model, the best hyperparameter combination is selected through cross-validation to ensure the generalization ability of the model.

[0214] In some embodiments, the prediction device may use the parameter-optimized XGBoost model to predict the test set:

[0215] ;

[0216] in, represents the prediction results of the test set, Indicates the The prediction results of a decision tree, Represents the feature input of the test set.

[0217] The predicted results are then compared with the actual load data to analyze the accuracy of the model. The predicted results can also be displayed through visualization tools to help judge the performance of the model.

[0218] This embodiment collects historical load time series data of multiple regions in the power system, preprocesses the historical load time series data, obtains candidate load time series data, performs deep feature extraction on the candidate load time series data, obtains a deep feature matrix of each region, performs multidimensional feature analysis on the candidate load time series data, obtains a multidimensional feature matrix of each region, fuses the deep feature matrix with the multidimensional feature matrix to obtain target load features, and performs cluster analysis on the multiple regions based on the target load features to determine the regional clusters corresponding to each region, constructs a load prediction model corresponding to each regional cluster according to the target load features, and performs load prediction on the power system based on the load prediction model corresponding to each regional cluster. The system performs power load forecasting; since this embodiment performs deep feature analysis and multi-dimensional feature analysis on the load time series data of each region respectively, it can achieve in-depth mining of the load characteristics of each region, effectively avoiding the problem of ignoring the potential characteristics and differences of different regions, and improving the prediction accuracy. By fusing the deep features and multi-dimensional features, clustering analysis is performed based on the fused target load characteristics, clustering of the same type of regions is achieved, and a reasonable load forecasting model is provided for regions with the same characteristics, effectively improving the power load forecasting efficiency, more accurately capturing and reflecting the load change trends of each region, greatly improving the accuracy of the power system load forecast, and providing accurate forecast result support for power resource scheduling.

[0219] In addition, an embodiment of the present invention also proposes a computer-readable storage medium, which stores a multi-region power load prediction program. When the multi-region power load prediction program is executed by a processor, it implements the steps of the multi-region power load prediction method described above.

[0220] The computer-readable storage medium provided herein may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems, or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including, but not limited to, wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0221] The computer-readable storage medium may be included in the multi-region power load forecasting device; or may exist independently without being incorporated into the multi-region power load forecasting device.

[0222] In addition, an embodiment of the present invention further proposes a computer program product, including a multi-region power load forecasting program, which implements the steps of the multi-region power load forecasting method described above when executed by a processor.

[0223] The specific implementation of the computer program product of the present invention is basically the same as the various embodiments of the above-mentioned multi-region power load forecasting method, and will not be repeated here.

[0224] Reference Figure 6 , Figure 6 This is a structural block diagram of the first embodiment of the multi-region power load forecasting device of the present invention.

[0225] like Figure 6 As shown, the multi-region power load forecasting device proposed in the embodiment of the present invention includes:

[0226] The data processing module 10 is used to collect historical load time series data of multiple areas in the power system and pre-process the historical load time series data to obtain candidate load time series data;

[0227] A deep feature analysis module 20 is used to extract deep features from the candidate load time series data to obtain a deep feature matrix for each region;

[0228] A multidimensional feature analysis module 30 is used to perform multidimensional feature analysis on the candidate load time series data to obtain a multidimensional feature matrix for each region;

[0229] a feature clustering analysis module 40 for performing feature fusion on the deep feature matrix and the multidimensional feature matrix to obtain target load features, and performing cluster analysis on the multiple regions based on the target load features to determine a region cluster corresponding to each region;

[0230] A prediction model building module 50 is used to build a load prediction model corresponding to each regional cluster according to the target load characteristics;

[0231] The power load forecasting module 60 is configured to forecast the power load of the power system based on the load forecasting model corresponding to each regional cluster.

[0232] This embodiment collects historical load time series data of multiple regions in the power system, preprocesses the historical load time series data, obtains candidate load time series data, performs deep feature extraction on the candidate load time series data, obtains a deep feature matrix of each region, performs multidimensional feature analysis on the candidate load time series data, obtains a multidimensional feature matrix of each region, fuses the deep feature matrix with the multidimensional feature matrix to obtain target load features, and performs cluster analysis on the multiple regions based on the target load features to determine the regional clusters corresponding to each region, constructs a load prediction model corresponding to each regional cluster according to the target load features, and performs load prediction on the power system based on the load prediction model corresponding to each regional cluster. The system performs power load forecasting; since this embodiment performs deep feature analysis and multi-dimensional feature analysis on the load time series data of each region respectively, it can achieve in-depth mining of the load characteristics of each region, effectively avoiding the problem of ignoring the potential characteristics and differences of different regions, and improving the prediction accuracy. By fusing the deep features and multi-dimensional features, clustering analysis is performed based on the fused target load characteristics, clustering of the same type of regions is achieved, and a reasonable load forecasting model is provided for regions with the same characteristics, effectively improving the power load forecasting efficiency, more accurately capturing and reflecting the load change trends of each region, greatly improving the accuracy of the power system load forecast, and providing accurate forecast result support for power resource scheduling.

[0233] The multi-region power load forecasting device provided in this application utilizes the multi-region power load forecasting method described in the aforementioned embodiments to address the technical issues of multi-region power load forecasting. Compared to the prior art, the multi-region power load forecasting device provided in this application offers the same beneficial effects as the multi-region power load forecasting method described in the aforementioned embodiments. Other technical features of the multi-region power load forecasting device are the same as those disclosed in the aforementioned embodiments and are not further elaborated upon here.

[0234] It should be understood that the above is only an example and does not constitute any limitation to the technical solution of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any limitation on this.

[0235] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present invention. In practical applications, technicians in this field can select part or all of it according to actual needs to achieve the purpose of the embodiment scheme, and no limitation is made here.

[0236] In addition, for technical details not fully described in this embodiment, please refer to the multi-region power load forecasting method provided in any embodiment of the present invention, and will not be repeated here.

[0237] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or system comprising the element.

[0238] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0239] Through the above description of the embodiments, those skilled in the art will clearly understand that the above-mentioned embodiments and methods can be implemented by means of software plus the necessary general-purpose hardware platform. Of course, hardware can also be used, but in many cases the former is a more preferred embodiment. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory / random access memory, a magnetic disk, or an optical disk) and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0240] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A multi-region power load forecasting method, characterized in that: The multi-region power load forecasting method includes: Collecting historical load time series data of multiple areas in the power system, and preprocessing the historical load time series data to obtain candidate load time series data; Performing deep feature extraction on the candidate load time series data to obtain a deep feature matrix for each region; Performing multi-dimensional feature analysis on the candidate load time series data to obtain a multi-dimensional feature matrix for each region; Performing feature fusion on the deep feature matrix and the multidimensional feature matrix to obtain target load features, and performing cluster analysis on the multiple regions based on the target load features to determine a region cluster corresponding to each region; Constructing a load forecasting model corresponding to each regional cluster according to the target load characteristics; Performing power load forecasting on the power system based on the load forecasting model corresponding to each regional cluster; The performing multi-dimensional feature analysis on the candidate load time series data to obtain a multi-dimensional feature matrix of each region includes: Calculate the mean, standard deviation, and coefficient of variation of the candidate load time series data: in, represents the mean of candidate load time series data, represents the standard deviation of the candidate load time series data, represents the coefficient of variation; Performing global load fluctuation analysis based on the mean, the standard deviation, and the coefficient of variation to obtain basic statistical characteristics; Perform time fluctuation analysis based on the candidate load time series data to obtain time pattern characteristics, which include peak-to-valley difference, time period average ratio, and daily load rate: in, Indicates the peak-to-valley difference of load within a day. Indicates the maximum load in a day. Indicates the minimum load value in a day. It represents the load average ratio between morning and evening hours. represents the average load value during the morning period, Indicates the average load value during the night time. represents the daily load rate, Indicates the average load value within a day; A short-term fluctuation analysis is performed based on the candidate load time series data to obtain a volatility feature, wherein the volatility feature includes a lagged autocorrelation coefficient: in, represents the lagged autocorrelation coefficient, Indicates the The load value of a time step is a data point in the candidate load time series data. represents the total number of time steps, represents the number of lag steps, represents the data mean of the candidate load time series data; The basic statistical features, the time pattern features and the volatility features are combined into a multi-dimensional feature matrix.

2. The multi-region power load forecasting method according to claim 1, wherein: The step of performing deep feature extraction on the candidate load time series data to obtain a deep feature matrix of each region includes: Performing a bidirectional long-term and short-term dependency analysis on the candidate load time series data to obtain a forward hidden state and a reverse hidden state of each time step, and concatenating the forward hidden state with the reverse hidden state to obtain a bidirectional dependency feature of each time step; Performing linear transformation based on the bidirectional hidden state to construct a query matrix, a key matrix, and a value matrix; Determine the attention features for each time step based on the query matrix, the key matrix, and the value matrix: in, Represents the time step The attention characteristics of represents the query matrix, represents the matrix, represents the value matrix, represents transpose, represents the dimension of the bond matrix, represents the activation function; Concatenate the bidirectional dependency features and attention features of each time step to obtain the initial deep features: in, Represents the time step The initial depth features of represents the feature concatenation operation, Represents the time step Bidirectional dependency characteristics; Construct a deep feature matrix based on the initial deep features of each time step: in, represents the deep feature matrix, represents the time dimension, Represents the feature dimension.

3. The multi-region power load forecasting method according to claim 2, wherein: The step of fusing the deep feature matrix with the multidimensional feature matrix to obtain target load features, and performing cluster analysis on the multiple regions based on the target load features to determine region clusters corresponding to the respective regions includes: The deep feature matrix and the multi-dimensional feature matrix are subjected to feature fusion to obtain the target load feature: in, represents the target load characteristics, Represents a multidimensional feature matrix; Determine the feature similarity between regions based on the target load feature: in, Indicates area and region The feature similarity between The characteristic dimension representing the target load characteristics, Indicates area No. Features, Indicates area No. Features Cluster analysis is performed on the multiple regions according to the feature similarity to determine a region cluster corresponding to each region.

4. The multi-region power load forecasting method according to claim 3, wherein: The constructing of a load forecasting model corresponding to each regional cluster according to the target load characteristics includes: Construct an initial prediction model corresponding to each regional cluster. The objective function of the initial prediction model is as follows: in, represents the objective function, represents the loss function, represents the number of samples, Indicates the The true value of the sample, Indicates the Samples passed The predicted value after the decision tree, represents a set of decision trees, represents the regularization term used to prevent overfitting, Indicates the The complexity function of a decision tree; The target load characteristics are normalized: in, Represents the normalized The first sample Features, represents the first The first sample Features, and Respectively represent The maximum and minimum values ​​of the features; The importance of each feature in the normalized target load feature is evaluated, and the normalized target load feature is screened based on the importance evaluation result to obtain the target training feature. The importance evaluation refers to the following formula: in, Represents a decision tree The gain value of the feature in , Indicates the result of materiality assessment; The initial prediction model is trained based on the target training features to obtain a load prediction model.

5. The multi-region power load forecasting method according to any one of claims 1 to 4, characterized in that: The preprocessing of the historical load time series data to obtain candidate load time series data includes: Calculate the standard score for each data point in the historical load time series data: in, Represents a data point The standard score of represents the mean of the historical load time series data, Represents the standard deviation of historical load time series data; Marking data points with standard scores higher than an abnormal threshold in the historical load time series data as abnormal data points, and performing data cleaning on the abnormal data points; Fill missing data in the historical load time series data after data cleaning, obtain the initial load time series data, and fill missing data according to the following formula: in, Indicates the length of the data missing segment, Indicates the missing position of the current missing value in the missing segment, represents the known data points after the missing segment, represents the known data points before the missing segment; The initial load time series data is normalized to obtain candidate load time series data. The normalization process is performed according to the following formula: in, represents the normalized data points, represents the minimum data point in the initial load time series data, Indicates the maximum value data point.

6. A multi-region power load forecasting device, characterized in that: The multi-region power load forecasting device includes: A data processing module is used to collect historical load time series data of multiple areas in the power system and preprocess the historical load time series data to obtain candidate load time series data; A deep feature analysis module is used to extract deep features from the candidate load time series data to obtain a deep feature matrix for each region; A multidimensional feature analysis module is used to perform multidimensional feature analysis on the candidate load time series data to obtain a multidimensional feature matrix for each region; a feature clustering analysis module, configured to perform feature fusion on the deep feature matrix and the multidimensional feature matrix to obtain target load features, and to perform cluster analysis on the multiple regions based on the target load features to determine a region cluster corresponding to each region; A prediction model building module, configured to build a load prediction model corresponding to each regional cluster according to the target load characteristics; An electric power load forecasting module, configured to forecast the electric power load of the electric power system based on the load forecasting model corresponding to each regional cluster; The multidimensional feature analysis module is further used to calculate the mean, standard deviation and coefficient of variation of the candidate load time series data: in, represents the mean of candidate load time series data, represents the standard deviation of the candidate load time series data, represents the coefficient of variation; Perform global load fluctuation analysis based on the mean, standard deviation, and coefficient of variation to obtain basic statistical features; perform time fluctuation analysis based on the candidate load time series data to obtain time pattern features, which include peak-to-valley difference, time period mean ratio, and daily load rate: in, Indicates the peak-to-valley difference of load within a day. Indicates the maximum load in a day. Indicates the minimum load value in a day. It represents the load average ratio between morning and evening hours. represents the average load value during the morning period, Indicates the average load value during the night time. represents the daily load rate, Indicates the average load value within a day; A short-term fluctuation analysis is performed based on the candidate load time series data to obtain a volatility feature, wherein the volatility feature includes a lagged autocorrelation coefficient: in, represents the lagged autocorrelation coefficient, Indicates the The load value of a time step is a data point in the candidate load time series data. represents the total number of time steps, represents the number of lag steps, represents the data mean of the candidate load time series data; The basic statistical features, the time pattern features and the volatility features are combined into a multi-dimensional feature matrix.

7. A multi-region power load forecasting device, characterized in that: The multi-region power load forecasting device includes: a memory, a processor, and a multi-region power load forecasting program stored on the memory and executable on the processor, wherein the multi-region power load forecasting program is configured to implement the multi-region power load forecasting method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a multi-region power load prediction program, which, when executed by a processor, implements the multi-region power load prediction method according to any one of claims 1 to 5.

9. A computer program product, characterized in that The computer program product includes a multi-region power load forecasting program, which implements the steps of the multi-region power load forecasting method according to any one of claims 1 to 5 when executed by a processor.

Images