Short-term power load interval prediction method based on kernel principal component regression analysis
By using a kernel principal component regression analysis method, historical power load data is clustered and dimensionality reduced to construct a load interval prediction model. This solves the problems of high model complexity and prediction uncertainty in traditional short-term power load prediction, and achieves efficient load interval prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-16
- Publication Date
- 2026-03-13
AI Technical Summary
Traditional short-term power load forecasting methods struggle to overcome the uncertainty of forecast results while maintaining forecast efficiency. They are also sensitive to multicollinearity of input variables, have complex model structures, and low training efficiency.
By employing a kernel principal component regression analysis method, clustering and dimensionality reduction of historical power load impact data are performed to extract the main power load impact characteristics, construct a load interval prediction model, and provide short-term power load interval prediction for the power system.
It reduces model complexity and computational cost, provides reliable load range prediction results, reflects the uncertainty of prediction values, and provides a more comprehensive reference for power grid dispatching decisions.
Smart Images
Figure CN121660185A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power system load forecasting technology, and in particular to a short-term power load interval forecasting method based on kernel principal component regression analysis. Background Technology
[0002] Short-term load forecasting is a crucial foundation for power system operation and planning. Accurate load forecasting helps power sectors rationally plan generation, improving the economy and reliability of grid operation. With the large-scale integration of renewable energy into the grid, the uncertainty facing the power system is increasing, and traditional point-based forecasting methods struggle to fully reflect the uncertainty characteristics of the load.
[0003] Currently, short-term load forecasting methods are mainly divided into two categories: classical forecasting methods (such as time series analysis and regression analysis) and modern intelligent forecasting methods (such as artificial neural networks and support vector machines). However, these methods generally have the following limitations: point forecasting methods can only provide a single forecast value and cannot quantify the uncertainty of load forecasting; most methods are sensitive to multicollinearity among input variables; and when there are too many input influencing factors, the model structure becomes complex and training efficiency is low.
[0004] In recent years, interval forecasting has attracted widespread attention as a method to quantify load uncertainty. For example, some studies have combined quantile regression with deep learning for load interval forecasting. However, these methods are usually complex and computationally expensive.
[0005] Therefore, developing a short-term load forecasting method that can reduce model complexity and computational cost while providing a reliable forecast range has significant theoretical and practical value. Summary of the Invention
[0006] This invention provides a short-term power load interval prediction method based on kernel principal component regression analysis to solve the problem that traditional short-term power load prediction is difficult to overcome the uncertainty of prediction results while taking into account prediction efficiency.
[0007] In a first aspect, embodiments of the present invention provide a short-term power load interval prediction method based on kernel principal component regression analysis, including: Obtain historical power load data and corresponding historical power load impact data; The historical power load impact data is clustered, and the corresponding historical power load impact data for each power consumption scenario is determined according to the clustering results. For the historical power load impact data corresponding to each power consumption scenario, dimensionality reduction is performed based on kernel principal component analysis to extract the main power load impact features corresponding to each power consumption scenario; A load range prediction model is constructed based on the main power load impact characteristics corresponding to each power consumption scenario and the corresponding historical power load data. Based on the load interval prediction model corresponding to each electricity consumption scenario, the short-term power load of the power system is predicted within the interval.
[0008] Secondly, embodiments of the present invention provide a short-term power load interval prediction device based on kernel principal component regression analysis, comprising: The acquisition module is used to acquire historical power load data and corresponding historical power load impact data; The clustering module is used to cluster the historical power load impact data and determine the corresponding historical power load impact data for each power consumption scenario based on the clustering results. The kernel principal component analysis module is used to perform dimensionality reduction based on the historical power load impact data for each power consumption scenario, and extract the main power load impact features for each power consumption scenario. The prediction model building module is used to build a load interval prediction model based on the main power load impact characteristics corresponding to each power consumption scenario and the corresponding historical power load data. The load forecasting module is used to perform interval forecasting of the short-term power load of the power system based on the load interval forecasting model corresponding to each power consumption scenario.
[0009] Thirdly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect or any possible implementation thereof.
[0010] In this embodiment of the invention, historical power load data and corresponding historical power load impact data are acquired. First, the historical power load impact data is clustered. Based on the clustering results, the historical power load impact data corresponding to each power consumption scenario is determined. Then, for the historical power load impact data corresponding to each power consumption scenario, dimensionality reduction is performed based on kernel principal component analysis to extract the main power load impact features corresponding to each power consumption scenario. Thus, clustering and kernel principal component analysis distinguish the main power load impact features of different power consumption scenarios. Based on this, a load interval prediction model is constructed based on the main power load impact features corresponding to each power consumption scenario and the corresponding historical power load data. Then, based on the load interval prediction model corresponding to each power consumption scenario, interval prediction of the short-term power load of the power system is performed. This fully considers the nonlinear characteristics of data under different power consumption scenarios, provides interval prediction results for short-term power load prediction of the power system, reflects the uncertainty of the predicted value, and provides a more comprehensive reference for power grid dispatching decisions. Attached Figure Description
[0011] Figure 1 This is a flowchart illustrating the implementation of the short-term power load interval prediction method based on kernel principal component regression analysis provided in this embodiment of the invention. Figure 2 This is a schematic diagram of the short-term power load interval prediction device based on kernel principal component regression analysis provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation
[0012] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0013] See Figure 1 The document illustrates a flowchart of the implementation of the short-term power load interval prediction method based on kernel principal component regression analysis provided in this embodiment of the invention, detailed below: Step 101: Obtain historical power load data and corresponding historical power load impact data.
[0014] Historical power load data refers to power load data for historical periods, while historical power load impact data refers to data that may affect power load data during historical periods, such as meteorological data (e.g., temperature, humidity, wind speed) and date-type data (e.g., weekdays, holidays).
[0015] For example, after obtaining historical power load data and corresponding historical power load impact data, the process may further include: preprocessing the historical power load data and corresponding historical power load impact data.
[0016] Preprocessing can include outlier handling, missing value imputation, and normalization. By preprocessing historical power load data and corresponding historical power load impact data, the impact of outliers and data formats on subsequent power load forecasting can be avoided.
[0017] Step 102: Cluster the historical power load impact data and determine the corresponding historical power load impact data for each power consumption scenario based on the clustering results.
[0018] In this embodiment, considering that different influencing factors may affect the power load in different scenarios, before extracting the main components affecting the power load based on kernel principal component analysis, the historical power load impact data is first clustered. According to the clustering results, the historical power load impact data is divided into different categories, each category corresponding to a power consumption scenario, so that the main components affecting the power load can be extracted for different power consumption scenarios in the future.
[0019] In one embodiment, step 102 includes: Clustering is performed on historical power load impact data to obtain clustering results.
[0020] Each clustering result is treated as an electricity consumption scenario, and historical power load impact data corresponding to each electricity consumption scenario is obtained.
[0021] For example, considering that it is not easy to know in advance which categories the historical power load impact data can be divided into, we can consider using methods such as hierarchical clustering and DBSCAN clustering to cluster the historical power load impact data to obtain clustering results.
[0022] Step 103: For the historical power load impact data corresponding to each power consumption scenario, perform dimensionality reduction based on kernel principal component analysis to extract the main power load impact features corresponding to each power consumption scenario.
[0023] In one embodiment, step 103 includes: Historical power load impact data for each power consumption scenario are used as raw data, and the raw data are mapped to a high-dimensional feature space based on a preset kernel function.
[0024] Principal component analysis is performed in a high-dimensional feature space to calculate the eigenvalues and eigenvectors of the kernel matrix.
[0025] Select the eigenvectors corresponding to the top k largest eigenvalues as the principal component directions, so that the cumulative contribution rate of the eigenvalues reaches a preset threshold.
[0026] By projecting the raw data onto the principal component direction, the main power load impact characteristics corresponding to each power consumption scenario are obtained.
[0027] In this embodiment, kernel principal component analysis is used to perform nonlinear dimensionality reduction on the clustered multidimensional input variables to extract the main feature components (i.e., the main power load impact features). This effectively reduces the dimensionality of the input variables and improves model training efficiency and generalization ability. For example, the specific process may include: Choose an appropriate preset kernel function (such as the Gaussian radial basis kernel function) to map the original data to a high-dimensional feature space.
[0028] Principal component analysis is performed in the feature space to calculate the eigenvalues and eigenvectors of the kernel matrix.
[0029] The number of principal components is determined based on the cumulative contribution rate, and the main feature components are extracted.
[0030] For example, solve for the eigenvalues and eigenvectors of the kernel matrix, select the eigenvectors corresponding to the top k largest eigenvalues as the principal component directions, so that the cumulative contribution rate of the eigenvalues reaches a preset threshold (such as 85%), and then project the original data onto the principal component directions to obtain the dimensionality-reduced features, which are the main power load impact features corresponding to each power consumption scenario.
[0031] Step 104: Construct a load range prediction model based on the main power load impact characteristics and corresponding historical power load data for each power consumption scenario.
[0032] For example, step 104 includes: Based on the main power load impact characteristics and corresponding historical power load data for each power consumption scenario, a quantile regression model is constructed as a load interval prediction model.
[0033] In this embodiment, after obtaining the dimensionality-reduced features, a quantile regression model is constructed for each electricity consumption scenario. Quantile regression can directly estimate the conditional quantiles without making any assumptions about the error distribution.
[0034] For a given quantile The quantile regression model can be expressed as: ; in, This is the load forecast value. The input feature vector represents the main power load impact characteristics corresponding to each power consumption scenario. quantiles The corresponding regression coefficients.
[0035] The following loss function can be optimized to estimate : ; in, This represents the actual load value.
[0036] Step 105: Perform short-term power load forecasting for the power system based on the load range forecasting model corresponding to each power consumption scenario.
[0037] In one embodiment, step 105 includes: Obtain data on the impact of power load on the period to be predicted.
[0038] Based on the power load impact data, the target load interval prediction model is determined from the load interval prediction model corresponding to each power consumption scenario.
[0039] Based on power load impact data and target load range prediction models, short-term power load of the power system is predicted within ranges.
[0040] In this embodiment, since a corresponding load interval prediction model is constructed for each electricity consumption scenario, when predicting the power load, the target load interval prediction model can be determined from the load interval prediction model corresponding to each electricity consumption scenario based on the power load impact data. Then, the short-term power load of the power system can be predicted based on the power load impact data and the target load interval prediction model.
[0041] For example, the target load range prediction model is determined from the load range prediction model corresponding to each electricity consumption scenario based on the power load impact data, including: Calculate the distance between the power load impact data and the corresponding historical power load impact data for each power consumption scenario.
[0042] The scenario with the smallest distance is identified as the target scenario, and the load interval prediction model corresponding to the target scenario is identified as the target load interval prediction model.
[0043] For example, calculating the distance between the power load impact data and the corresponding historical power load impact data for each power consumption scenario includes: Calculate the distance between the power load impact data and each cluster center in the clustering results, and use this distance as the distance between the power load impact data and the historical power load impact data corresponding to each power consumption scenario.
[0044] In this embodiment, considering that multiple cluster centers can be obtained when clustering the historical power load impact data, and each cluster center can represent a power consumption scenario, the distance between the power load impact data and each cluster center in the clustering result can be calculated as the distance between the power load impact data and the historical power load impact data corresponding to each power consumption scenario.
[0045] This invention acquires historical power load data and corresponding historical power load impact data. First, it clusters the historical power load impact data, determining the historical power load impact data for each power consumption scenario based on the clustering results. Then, for the historical power load impact data corresponding to each power consumption scenario, it performs dimensionality reduction based on kernel principal component analysis (KPI) to extract the main power load impact features corresponding to each power consumption scenario. Thus, clustering and KPI distinguish the main power load impact features of different power consumption scenarios. Based on this, a load interval prediction model is constructed based on the main power load impact features corresponding to each power consumption scenario and the corresponding historical power load data. Then, based on the load interval prediction model corresponding to each power consumption scenario, interval prediction of the short-term power load of the power system is performed. This fully considers the nonlinear characteristics of data under different power consumption scenarios, providing interval results for short-term power load prediction of the power system, reflecting the uncertainty of the predicted values, and providing a more comprehensive reference for power grid dispatching decisions.
[0046] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
[0047] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.
[0048] Figure 2 The diagram shows a schematic of a short-term power load interval prediction device based on kernel principal component regression analysis provided in an embodiment of the present invention. For ease of explanation, only the parts relevant to the embodiment of the present invention are shown, and are described in detail below: like Figure 2 As shown, the short-term power load interval forecasting device based on kernel principal component regression analysis includes: The acquisition module 21 is used to acquire historical power load data and corresponding historical power load impact data.
[0049] Clustering module 22 is used to cluster historical power load impact data and determine the corresponding historical power load impact data for each power consumption scenario based on the clustering results.
[0050] The kernel principal component analysis module 23 is used to perform dimensionality reduction based on the historical power load impact data corresponding to each power consumption scenario, and extract the main power load impact features corresponding to each power consumption scenario.
[0051] The prediction model building module 24 is used to build a load range prediction model based on the main power load impact characteristics and corresponding historical power load data for each power consumption scenario.
[0052] The load forecasting module 25 is used to forecast the short-term power load of the power system according to the load interval forecasting model corresponding to each power consumption scenario.
[0053] This invention acquires historical power load data and corresponding historical power load impact data. First, it clusters the historical power load impact data, determining the historical power load impact data for each power consumption scenario based on the clustering results. Then, for the historical power load impact data corresponding to each power consumption scenario, it performs dimensionality reduction based on kernel principal component analysis (KPI) to extract the main power load impact features corresponding to each power consumption scenario. Thus, clustering and KPI distinguish the main power load impact features of different power consumption scenarios. Based on this, a load interval prediction model is constructed based on the main power load impact features corresponding to each power consumption scenario and the corresponding historical power load data. Then, based on the load interval prediction model corresponding to each power consumption scenario, interval prediction of the short-term power load of the power system is performed. This fully considers the nonlinear characteristics of data under different power consumption scenarios, providing interval results for short-term power load prediction of the power system, reflecting the uncertainty of the predicted values, and providing a more comprehensive reference for power grid dispatching decisions.
[0054] In one possible implementation, clustering module 22 is specifically used for: Clustering is performed on historical power load impact data to obtain clustering results.
[0055] Each clustering result is treated as an electricity consumption scenario, and historical power load impact data corresponding to each electricity consumption scenario is obtained.
[0056] In one possible implementation, the kernel principal component analysis module 23 is specifically used for: Historical power load impact data for each power consumption scenario are used as raw data, and the raw data are mapped to a high-dimensional feature space based on a preset kernel function.
[0057] Principal component analysis is performed in a high-dimensional feature space to calculate the eigenvalues and eigenvectors of the kernel matrix.
[0058] Select the eigenvectors corresponding to the top k largest eigenvalues as the principal component directions, so that the cumulative contribution rate of the eigenvalues reaches a preset threshold.
[0059] By projecting the raw data onto the principal component direction, the main power load impact characteristics corresponding to each power consumption scenario are obtained.
[0060] In one possible implementation, the prediction model building module 24 is specifically used for: Based on the main power load impact characteristics and corresponding historical power load data for each power consumption scenario, a quantile regression model is constructed as a load interval prediction model.
[0061] In one possible implementation, the load forecasting module 25 is specifically used for: Obtain data on the impact of power load on the period to be predicted.
[0062] Based on the power load impact data, the target load interval prediction model is determined from the load interval prediction model corresponding to each power consumption scenario.
[0063] Based on power load impact data and target load range prediction models, short-term power load of the power system is predicted within ranges.
[0064] In one possible implementation, the load forecasting module 25 is specifically used for: Calculate the distance between the power load impact data and the corresponding historical power load impact data for each power consumption scenario.
[0065] The scenario with the smallest distance is identified as the target scenario, and the load interval prediction model corresponding to the target scenario is identified as the target load interval prediction model.
[0066] In one possible implementation, the load forecasting module 25 is specifically used for: Calculate the distance between the power load impact data and each cluster center in the clustering results, and use this distance as the distance between the power load impact data and the historical power load impact data corresponding to each power consumption scenario.
[0067] In one possible implementation, module 21 is also used for: Preprocessing is performed on historical power load data and corresponding historical power load impact data.
[0068] Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. For example... Figure 3 As shown, the electronic device 3 of this embodiment includes a processor 30 and a memory 31. The memory 31 stores a computer program 32. When the processor 30 executes the computer program 32, it implements the steps in the various method embodiments described above. Alternatively, when the processor 30 executes the computer program 32, it implements the functions of each module / unit in the various device embodiments described above.
[0069] For example, computer program 32 may be divided into one or more modules / units, which are stored in memory 31 and executed by processor 30 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 32 in electronic device 3.
[0070] Electronic device 3 may include, but is not limited to, processor 30 and memory 31. Those skilled in the art will understand that... Figure 3This is merely an example of electronic device 3 and does not constitute a limitation on electronic device 3. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device 3 may also include input / output devices, network access devices, buses, etc.
[0071] For the sake of simplicity and clarity, only the above-described functional modules / units are used as examples. In practical applications, the functions described above can be assigned to different functional modules / units as needed. These modules / units can be implemented in hardware, software, or a combination of both.
[0072] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Unless otherwise specified or in conflict with logic, the terminology and / or descriptions between different embodiments are consistent and can be referenced interchangeably. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.
[0073] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A method for short-term power load interval forecasting based on kernel principal component regression analysis, characterized in that, include: Obtain historical power load data and corresponding historical power load impact data; The historical power load impact data is clustered, and the corresponding historical power load impact data for each power consumption scenario is determined according to the clustering results. For the historical power load impact data corresponding to each power consumption scenario, dimensionality reduction is performed based on kernel principal component analysis to extract the main power load impact features corresponding to each power consumption scenario; A load range prediction model is constructed based on the main power load impact characteristics corresponding to each power consumption scenario and the corresponding historical power load data. Based on the load interval prediction model corresponding to each electricity consumption scenario, the short-term power load of the power system is predicted within the interval.
2. The short-term power load interval prediction method based on kernel principal component regression analysis according to claim 1, characterized in that, The historical power load impact data is clustered, and the corresponding historical power load impact data for each power consumption scenario is determined according to the clustering results, including: Cluster the historical power load impact data to obtain clustering results; Each clustering result is treated as an electricity consumption scenario, and historical power load impact data corresponding to each electricity consumption scenario is obtained.
3. The short-term power load interval prediction method based on kernel principal component regression analysis according to claim 1, characterized in that, For the historical power load impact data corresponding to each power consumption scenario, dimensionality reduction is performed based on kernel principal component analysis to extract the main power load impact features corresponding to each power consumption scenario, including: The historical power load impact data corresponding to each power consumption scenario is used as the raw data, and the raw data is mapped to a high-dimensional feature space based on a preset kernel function. Principal component analysis is performed in the high-dimensional feature space to calculate the eigenvalues and eigenvectors of the kernel matrix. Select the eigenvectors corresponding to the top k largest eigenvalues as the principal component directions, so that the cumulative contribution rate of the eigenvalues reaches a preset threshold. The original data is projected onto the principal component direction to obtain the main power load impact characteristics corresponding to each power consumption scenario.
4. The short-term power load interval prediction method based on kernel principal component regression analysis according to claim 1, characterized in that, A load range prediction model is constructed based on the main power load impact characteristics corresponding to each power consumption scenario and the corresponding historical power load data, including: Based on the main power load impact characteristics corresponding to each power consumption scenario and the corresponding historical power load data, a quantile regression model is constructed as a load interval prediction model.
5. The short-term power load interval prediction method based on kernel principal component regression analysis according to claim 1, characterized in that, Based on the load interval prediction model corresponding to each electricity consumption scenario, the short-term power load of the power system is predicted in intervals, including: Obtain data on the impact of power load on the period to be predicted; Based on the power load impact data, a target load range prediction model is determined from the load range prediction model corresponding to each power consumption scenario; Based on the power load impact data and the target load interval prediction model, the short-term power load of the power system is predicted within an interval.
6. The short-term power load interval prediction method based on kernel principal component regression analysis according to claim 5, characterized in that, Based on the power load impact data, a target load interval prediction model is determined from the load interval prediction model corresponding to each power consumption scenario, including: Calculate the distance between the power load impact data and the corresponding historical power load impact data for each power consumption scenario; The scenario with the smallest distance is identified as the target scenario, and the load interval prediction model corresponding to the target scenario is identified as the target load interval prediction model.
7. The short-term power load interval prediction method based on kernel principal component regression analysis according to claim 6, characterized in that, Calculating the distance between the power load impact data and the corresponding historical power load impact data for each power consumption scenario includes: The distance between the power load impact data and each cluster center in the clustering results is calculated and used as the distance between the power load impact data and the historical power load impact data corresponding to each power consumption scenario.
8. The short-term power load interval prediction method based on kernel principal component regression analysis according to claim 1, characterized in that, After obtaining historical power load data and corresponding historical power load impact data, the following is also included: The historical power load data and the corresponding historical power load impact data are preprocessed.
9. A short-term power load interval prediction device based on kernel principal component regression analysis, characterized in that, include: The acquisition module is used to acquire historical power load data and corresponding historical power load impact data; The clustering module is used to cluster the historical power load impact data and determine the corresponding historical power load impact data for each power consumption scenario based on the clustering results. The kernel principal component analysis module is used to perform dimensionality reduction based on the historical power load impact data for each power consumption scenario, and extract the main power load impact features for each power consumption scenario. The prediction model building module is used to build a load interval prediction model based on the main power load impact characteristics corresponding to each power consumption scenario and the corresponding historical power load data. The load forecasting module is used to perform interval forecasting of the short-term power load of the power system based on the load interval forecasting model corresponding to each power consumption scenario.
10. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 8.