A method and system for predicting power load based on regional portrait

Abstract

The present invention discloses a method for predicting power load based on regional portraits, and relates to the field of power load prediction. At present, power load prediction ignores the influence of regional characteristics on power load, resulting in the accuracy and reliability of prediction results being limited. The present invention first obtains regional benchmark data by collecting and preprocessing relevant data of the target area. Then, key indicators reflecting regional characteristics are extracted through data mining technology to construct a regional portrait. Based on this regional portrait, features related to power load are extracted, and a power load prediction model is constructed in combination with historical power load data. Finally, the power load prediction model is used to predict the regional portrait information of the target area to obtain a more accurate power load prediction result. This technical solution can comprehensively consider various characteristics of the region, thereby improving the accuracy and reliability of power load prediction, and providing strong support for the optimized operation of the power system.

Description

Technical Field

[0001] The present invention relates to the field of power load forecasting, and in particular to a method and system for power load forecasting based on regional portraits. Background Art

[0002] With the accelerated urbanization process and rapid economic development, the importance of power load forecasting in energy management and power grid planning has become increasingly prominent. Power load forecasting is not only related to the stable operation of the power system, but also directly affects the rational use of energy and the realization of energy conservation and emission reduction goals. At present, due to differences in economic development, industrial structure, population distribution and other factors in different regions, their power load characteristics show obvious regional characteristics. Traditional power load forecasting methods often ignore these regional differences, use a unified forecasting model for processing, and ignore the impact of regional characteristics on power load, resulting in limited accuracy and reliability of the forecast results, making it difficult to meet the actual needs of various regions. Summary of the invention

[0003] The technical problem to be solved and the technical task proposed by the present invention are to improve and perfect the existing technical solutions and provide a power load forecasting method based on regional portraits to improve the accuracy and reliability of power load forecasting. To this end, the present invention adopts the following technical solutions.

[0004] A method for predicting power load based on regional profiling, comprising the steps of:

[0005] 1) Collecting regional relevant data of the target area and preprocessing the regional relevant data to obtain regional benchmark data;

[0006] 2) Based on the regional benchmark data, extract regional key indicators reflecting regional characteristics through data mining, and use the regional key indicators to construct a target regional profile;

[0007] 3) Extracting power load related features based on the target area portrait;

[0008] 4) Building a power load forecasting model based on the power load related characteristics and combining historical power load data;

[0009] 5) Collecting regional portrait information of the target area, and combining it with the power load forecasting model to perform forecasting, and obtain the power load forecasting result of the target area.

[0010] This technical solution collects relevant data of the target area and performs preprocessing, combines regional characteristics (such as economic, social, environmental and other factors) to predict power load, and comprehensively considers regional characteristics to more accurately reflect the actual demand of power load and improve the accuracy of prediction. In step 2, the key indicators are extracted by using data mining technology, which helps to extract features that have an important impact on prediction from complex regional data, and can discover potential and important regional features that may be ignored by traditional prediction methods. This technical solution can reflect the changes in regional characteristics in real time by constructing a target area portrait and extracting relevant features of power load, so that the prediction model can be dynamically adjusted during the power load prediction process to improve the timeliness and accuracy of the prediction results. This technical solution combines historical power load data and extracted power load-related features, and uses trends and laws in historical data to construct a prediction model. Such a combination can improve the reliability and prediction accuracy of the model. Constructing a target area portrait and combining it with the power load prediction model for prediction enables the prediction results to reflect the comprehensive impact of regional characteristics, thereby providing more comprehensive and accurate prediction results and helping the power system to carry out more effective resource planning and scheduling.

[0011] As a preferred technical means: Step 1) includes:

[0012] 101) Based on big data, collect regional relevant data of the target area;

[0013] 102) Performing data cleaning, normalization processing and missing value processing on the regional related data to obtain the regional benchmark data, wherein the regional benchmark data includes regional data in multiple fields.

[0014] By collecting regional data through big data technology, we can obtain information from multiple fields (such as economy, society, environment, etc.). This provides comprehensive basic data for power load forecasting and ensures the breadth and representativeness of data coverage. Data cleaning, normalization and missing value processing can significantly improve the quality of data, ensure the accuracy and consistency of data, and reduce prediction errors caused by data errors or inconsistencies.

[0015] As a preferred technical means: in step 2), based on the regional benchmark data, regional key indicators reflecting regional characteristics are extracted through data mining, including:

[0016] 201) Determine the key indicator types based on the power transmission structure of the target area; the key indicator types include grid scale, transmission capacity, line length, voltage level and power supply reliability;

[0017] 202) extracting indicators by principal component analysis in combination with the key indicator types to obtain multiple regional key indicators;

[0018] Calculate the covariance matrix of regional benchmark data, which reflects the correlation between various indicators;

[0019] Use mathematical methods to solve the eigenvalues ​​and eigenvectors of the covariance matrix. The eigenvector represents the main direction of change of the data.

[0020] According to the size of the eigenvalues, the number of principal components is determined, and the explanatory power of the selected principal components on the regional benchmark data is evaluated to ensure that the selected principal components can fully represent the key features of the regional benchmark data;

[0021] The regional benchmark data is projected onto the selected principal component to obtain the coordinates of each sample in the new principal component space, which are the regional key indicators extracted in combination with the key indicator type.

[0022] This technical solution determines the key indicator type according to the power transmission structure of the target area, so that the extracted key indicators are more in line with the actual power demand and characteristics of the region, thereby ensuring that the extracted indicators are directly related to the power load and enhancing the relevance and accuracy of the prediction model.

[0023] By extracting key indicators through principal component analysis (PCA), the most explanatory and important indicators can be extracted from multidimensional data, which helps to reduce dimensions, improve the efficiency of data analysis, and discover potential key features that affect power load. Principal component analysis can effectively reduce redundant information in the data, extract a small number of representative key indicators, help simplify the complexity of the model, and improve the stability and interpretability of the analysis results. By selecting and extracting regional key indicators of great significance, the interpretability of the model can be enhanced. By understanding which indicators have a significant impact on power load forecasting, it is helpful to better understand the reasons for changes in power demand and support decision-making and optimization. This technical solution can adapt to the characteristics of different regions, determine the indicator type by analyzing the power transmission structure of the target area, and ensure that the selected key indicators are suitable for the power load characteristics of a specific region. Principal component analysis can process complex data sets and extract useful information from them, improve the ability and efficiency of data analysis, and combine the analysis of the power transmission structure to further improve the accuracy of data mining. At different times or conditions, key indicators can be re-extracted based on new data and changes in the power transmission structure, so as to dynamically adjust the prediction model to adapt to new situations. By accurately extracting and selecting key indicators, more representative input features can be provided for the subsequent power load forecasting model, thereby improving the prediction accuracy and reliability of the model. In summary, the technical solution of extracting regional key indicators through data mining and principal component analysis in step 2 provides strong support for power load forecasting through accurate indicator selection, reducing redundant information, and improving model interpretability and analysis capabilities. It has significant technical advantages and practical application value.

[0024] As a preferred technical means: in step 2), a target area profile is constructed based on the key indicators of the area, including:

[0025] 203) Collect regional key indicator data, conduct historical change trend analysis and indicator correlation analysis on the regional key indicators, and obtain change trend analysis results and indicator correlation analysis results; the key indicator data include power grid scale data, transmission capacity data, line length data and voltage level data;

[0026] 204) Based on the above trend analysis results and indicator correlation analysis results, a basic framework for the target area portrait is constructed, including determining the main dimensions and sub-dimensions of the portrait; the main dimensions refer to key indicators that can comprehensively and systematically reflect the power transmission structure and development status of the target area, including power transmission scale and capacity, line construction and operation status; the sub-dimensions refer to the supplement and expansion of the main dimensions, which are used to more deeply describe and characterize the detailed characteristics and status of the target area in specific aspects, including the grid coverage area, the length of lines of each voltage level, the proportion of different types of lines and the line aging rate; the basic framework is a hierarchical structure, with the main dimensions as the top layer and the sub-dimensions as the bottom layer, so as to form a clear and orderly structure;

[0027] 205) According to the basic framework, the key indicator data of the region are classified and organized to obtain a portrait of the target region, which includes the static characteristics, development trends and potential advantages of the region.

[0028] This technical solution analyzes the historical change trend of regional key indicators, which can reveal the evolution law of power load and its related characteristics over time, help identify long-term trends, periodic changes and potential anomalies in the region, and provide historical reference for prediction. Through indicator correlation analysis, the interaction and influence relationship between different key indicators can be revealed, which helps to understand how each indicator works together to affect the power load and provide comprehensive input features for the model. Based on the results of change trend and correlation analysis, the basic framework of the target area portrait is constructed to ensure the clarity and logic of the portrait structure, making the regional characteristics more organized. Determining the main dimensions and sub-dimensions of the portrait can comprehensively display the regional characteristics from different levels and angles, and enhance the comprehensiveness and meticulousness of the regional portrait. Classifying and organizing the regional key indicator data according to the basic framework makes the data structure more standardized and clear, which helps to improve the readability and availability of the data and facilitate subsequent analysis and decision-making. Constructing a detailed target area portrait can provide more accurate regional characteristic information, help improve the accuracy of the power load forecasting model, and accurate portraits can better reflect the actual situation of regional power demand. The target area portrait provides detailed regional characteristic data, which provides a clear basis for decisions such as power dispatching and resource allocation. The construction method of the target area portrait supports dynamic updates and adjustments. With the addition of new data and information, the regional portrait can be continuously updated to adapt to changes in regional characteristics and new needs. By building a multi-dimensional portrait framework, regional characteristics can be analyzed from multiple levels and angles, which helps to more fully understand the complex factors of regional power load. The systematic construction of the target area portrait and the clear dimensional division can intuitively display the regional characteristics through visualization, which helps relevant personnel better understand and communicate the analysis results.

[0029] As a preferred technical means: Step 3) includes:

[0030] 301) Based on the power dispatch data of the target area, identify the factors affecting power load; the influencing factors include weather conditions, economic development, population growth trends and energy policy adjustments;

[0031] 302) Based on the power load influencing factors and indicator correlation analysis results, power load related features are extracted through the target area portrait.

[0032] This technical solution identifies the influencing factors of power load according to the power dispatching data of the target area, and can more accurately understand which factors directly or indirectly affect the power load. This identification can reveal the key driving factors that affect power demand and provide a clear basis for feature extraction. By combining the results of power load influencing factors and indicator correlation analysis, the impact of multiple factors on power load can be comprehensively considered, which helps to fully understand the determinants of power load and improve the accuracy and comprehensiveness of feature extraction. Based on the target area portrait, the power load-related features can be extracted, and the regional characteristics and data relationships reflected in the regional portrait can be used to ensure that the extracted features are closely related to the actual situation of the target area, thereby improving the effectiveness of feature extraction. By identifying influencing factors and extracting relevant features, accurate input data can be provided for the power load forecasting model, which helps to improve the performance and accuracy of the forecasting model, thereby better predicting the power load. Identifying influencing factors and extracting relevant features can provide data-driven decision support for the operation and management of the power system. Decision makers can optimize power dispatching and resource allocation based on this information and improve the operating efficiency of the system. With the update of power dispatching data and regional portrait information, the identified influencing factors and extracted features can be dynamically adjusted to cope with changes in power load and the emergence of new situations. Combining power dispatch data with regional portrait extraction features can support in-depth analysis of complex power systems and help reveal the complex relationships and potential patterns of power load changes.

[0033] As a preferred technical means: Step 4) includes:

[0034] 401) Collect historical power load data;

[0035] 402) Based on the time dimension, aligning the power load related features with the historical power load data to ensure that the time periods covered by the two are the same;

[0036] If the time ranges of the two are inconsistent, the data is trimmed or filled according to the actual situation to unify the time range; the power load-related characteristic data and the historical power load data are sorted according to the timestamps to ensure that the time sequence of the data is consistent; then, the timestamps of the power load-related characteristic data and the historical power load data are matched one by one;

[0037] If the timestamps match exactly, the corresponding data is output directly;

[0038] If there is not an exact match, use the closest time point for an approximate match;

[0039] 403) Based on a support vector machine, supervised training is performed on the power load related features and historical power load data to construct a power load prediction model;

[0040] The aligned data set is divided into a training set and a test set; the training set is used to train the power load forecasting model, and the test set is used to evaluate the prediction performance of the power load forecasting model;

[0041] Then, according to actual needs, select the corresponding kernel function and set the corresponding parameters to build the power load forecasting model; when the power load data presents a linear or approximately linear relationship, select the linear kernel function and set the penalty coefficient to build the model so that the model can intuitively show the linear relationship between the feature and the target;

[0042] Using the aligned power load related features and historical power load data in the training set, training the power load forecasting model;

[0043] During the training process, a time window length is set, and the training data is intercepted by the time window length to obtain a plurality of first training data; a constraint analysis is performed on the obtained plurality of first training data to determine whether these data have the characteristics of continuity and stability in this interval; if there are abnormal values ​​or mutation values, smoothing is performed, and a plurality of adjacent data are used to perform mean calculation to replace these abnormal values ​​or mutation values; the plurality of first training data after smoothing is input into the power load prediction model to train the model; the power load prediction model learns the mapping relationship between the power load related features and the power load data, and finds a hyperplane that can maximize the classification interval;

[0044] Use the data in the test set to evaluate the performance of the trained power load forecasting model, and compare the model output results with the expected standards. If the expected standards are not met, reselect the time window length, divide new training data, simulate different patterns that may appear in different time periods, and repeat the training process.

[0045] If the expected effect is still not met after multiple iterations, the kernel function is reselected, the corresponding parameters are configured, and the above data division and training process is repeated until the expected effect is met; conversely, if the expected standard is met, the trained model is directly output to obtain the power load prediction model.

[0046] The collection of historical power load data provides a true record of power load changes. These data provide a solid foundation for building a prediction model and help capture the historical patterns and trends of power load. Aligning power load-related features with historical power load data based on the time dimension can ensure the synchronization of features and load data, improve the accuracy of the model and the relevance of predictions. Using support vector machines (SVMs) for supervised training can effectively handle complex nonlinear relationships. SVMs can find the best prediction boundary in the feature space by maximizing the classification boundary, thereby improving the accuracy and robustness of the prediction model. SVMs are suitable for processing nonlinear data relationships. By mapping data to high-dimensional space through kernel functions, they can handle complex power load data features and improve the accuracy of capturing and predicting complex load patterns. Supervised training based on historical data and power load-related features can make full use of labeled data for model optimization, thereby improving the accuracy and reliability of the model's prediction of future loads. The SVM model has good generalization ability and can adapt to the power load prediction needs of different regions and time periods. It can be adjusted and optimized according to new data to cope with different power load conditions. By aligning the power load-related features with historical data, feature selection and optimization can be effectively performed, ensuring that the features used in model training have high predictive relevance and improving the performance of the model. The SVM model has a clear structure and can provide a clear explanation for power load forecasting. Through support vectors and classification boundaries, it is possible to understand which features have a major impact on the forecast results. The SVM model can be retrained and adjusted based on new power load data. It has strong adaptability, supports dynamic response to changes in power load patterns, and improves the long-term forecasting ability of the model. When processing small and medium-sized data sets, SVM has high computational efficiency and can quickly perform model training and forecasting.

[0047] As a preferred technical means: Step 5) includes:

[0048] 501) Collecting regional image information of the target area according to a preset period through the data collection window;

[0049] 502) Extracting power load related characteristic data based on the regional portrait information;

[0050] 503) Using the power load related characteristic data and the power load prediction model to perform prediction, and obtain a power load prediction result.

[0051] This technical solution collects regional portrait information of the target area according to a preset period through a data collection window, which can ensure that the power load forecasting model uses the latest regional data, and real-time updates improve the timeliness and accuracy of the forecast. Systematically collecting regional portrait information according to a preset period helps to maintain the consistency and integrity of the data and avoid the randomness and incompleteness of data collection. Using the latest regional portrait information to extract the power load-related feature data can ensure that the extracted features accurately reflect the actual situation of the current area and improve the adaptability of the forecasting model to the current power load. By combining the extracted power load-related feature data with the power load forecasting model, the model can be effectively applied to predict new data, which improves the relevance and accuracy of the forecast results. Using the latest power load-related feature data for forecasting can more accurately reflect the actual situation of the current power load and help optimize the operation and management of the power system. Since the regional portrait information and power load-related feature data are constantly updated, the forecasting model can be dynamically adjusted and optimized to cope with changes and emergencies in the power load and improve the flexibility and responsiveness of the system. Through periodic updates and accurate forecasts, more accurate power load data support is provided to help decision makers make more reasonable and effective power dispatching and management decisions. Regular data collection and processing enhances the reliability of the forecasting system, reduces forecasting errors caused by outdated or inaccurate data, and improves the stability of the system. This technical solution is widely used in different target areas. Through periodic data collection and real-time updates, it adapts to various power load forecasting needs and improves the versatility and applicability of the model.

[0052] Another object of the present invention is to provide a power load forecasting system based on regional profiling. The power load forecasting system comprises:

[0053] Data processing module: used to collect regional relevant data of the target area and pre-process these data to obtain regional benchmark data;

[0054] Data mining module: Based on regional benchmark data, extract regional key indicators that reflect regional characteristics through data mining, and use these key indicators to build a target regional profile;

[0055] Feature extraction module: extracts power load related features based on the target area portrait;

[0056] Model building module: constructs a power load forecasting model based on power load-related characteristics and historical power load data;

[0057] Result prediction module: collects regional portrait information of the target area, and combines it with the power load prediction model to make predictions to obtain the power load prediction results of the target area.

[0058] Data processing module: Through systematic data collection and preprocessing, the quality and integrity of regional benchmark data are ensured. This approach improves the consistency and accuracy of data, laying a solid foundation for subsequent analysis and prediction.

[0059] In this technical solution, the data mining module: through in-depth data mining, it is possible to extract key indicators that reflect regional characteristics. This method can reveal the potential rules in regional data, help build a more accurate regional portrait, and provide valuable insights.

[0060] Feature extraction module: Use the target area image to extract the relevant features of the power load, ensuring that the extracted features can accurately reflect the factors affecting the power load. This precise feature extraction helps improve the accuracy and robustness of the prediction model.

[0061] Model building module: By combining the relevant characteristics of power load and historical power load data, the forecasting model is built using advanced modeling technology. This method can make full use of historical data and characteristic information to improve the accuracy and reliability of the forecasting model.

[0062] Result prediction module: By periodically collecting regional profile information and combining it with the prediction model, it can achieve real-time prediction of power load. This real-time performance improves the ability to respond to changes in power load and supports more effective power management and scheduling.

[0063] The system can dynamically adjust and optimize the forecast model according to new regional data and power load characteristics. The dynamic adaptability supports the response to changes in power load and emergencies, improving the flexibility and response capabilities of the system. The system provides accurate power load forecast results, which can effectively support decision makers in power dispatch and resource allocation, help optimize the operation of the power system, and improve energy efficiency. Through systematic data processing and precise feature extraction, the system improves the reliability of the forecast, reduces the impact of human intervention, and ensures the stability and consistency of the forecast results.

[0064] As a preferred technical means: the data processing module further includes:

[0065] Data collection unit: used to connect to the data source of the target area through an interface and collect area-related data in real time or periodically;

[0066] Data preprocessing unit: used to clean, normalize and fill missing values ​​of the collected regional related data to generate regional benchmark data;

[0067] Data storage unit: used to store pre-processed regional benchmark data and support subsequent data mining and analysis operations;

[0068] The data mining module further comprises:

[0069] Key indicator type determination unit: determines the key indicator type according to the power transmission structure of the target area;

[0070] Principal component analysis unit: extracting indicators by principal component analysis and combining the key indicator types to obtain multiple regional key indicators;

[0071] Data analysis unit: used to collect regional key indicator data, and perform historical change trend analysis and indicator correlation analysis on the regional key indicators, to obtain change trend analysis results and indicator correlation analysis results;

[0072] Portrait construction unit: Based on the results of change trend analysis and indicator correlation analysis, the basic framework of the target area portrait is constructed, including determining the main dimensions and sub-dimensions of the portrait;

[0073] Data organization unit: classify and organize regional key indicator data according to the basic framework to obtain a target regional portrait;

[0074] The feature extraction module further comprises:

[0075] Power load feature identification unit: used to identify key factors affecting power load based on power dispatch data of the target area;

[0076] Feature extraction unit: used to extract power load related feature data based on power load influencing factors and regional portrait information;

[0077] Feature storage unit: used to store the extracted power load-related feature data for further analysis and training by the model building module.

[0078] In this technical solution, the data collection unit: supports real-time or regular data collection from the data source of the target area, which improves the flexibility and timeliness of data collection. The data collection frequency can be adjusted as needed to ensure the latest regional relevant data. Data preprocessing unit: includes data cleaning, normalization processing and missing value filling. These processing steps ensure the quality and consistency of the data and effectively reduce data noise and errors. Data storage unit: provides persistent storage of preprocessed data, supports subsequent data mining and analysis operations, and ensures the reliability and accessibility of data. Key indicator type determination unit: determines the key indicator type according to the power transmission structure, can effectively identify the key features related to the power load, and helps to build an accurate regional portrait. Principal component analysis unit: extracts indicators through principal component analysis, effectively reduces the dimension and extracts key regional indicators, and optimizes data processing and analysis. Data analysis unit: performs historical change trends and correlation analysis on regional key indicators, provides a scientific basis for building a target regional portrait, and improves the accuracy of the regional portrait. Portrait construction unit: constructs the basic framework of the regional portrait based on the analysis results, and determines the main dimensions and sub-dimensions, which helps to fully and accurately describe the regional characteristics. Data organization unit: classify and organize regional key indicator data to ensure the structuring and systematization of regional portraits for subsequent use and analysis. Power load feature identification unit: identify the key factors of power load through power dispatch data, accurately identify the main factors affecting the load, and improve the targeted nature of the prediction. Feature extraction unit: extract power load-related feature data based on the identified influencing factors and regional portrait information to ensure the accuracy and relevance of the features. Feature storage unit: stores the extracted power load-related feature data, provides a reliable data foundation for the model building module, and supports further analysis and training. This technical solution provides high-quality data and feature inputs through systematic data processing, effective data mining, and precise feature extraction, supports the construction of a more accurate and robust power load forecasting model, ensures reliable storage and systematic organization of data, and supports the processing and analysis of large-scale data.

[0079] As a preferred technical means: the model building module further includes:

[0080] Historical data integration unit: used to collect and integrate historical power load data and power load related characteristic data;

[0081] Feature alignment unit: used to align power load related features with historical power load data based on the time dimension to ensure data consistency;

[0082] Model training unit: used to conduct supervised training on power load related features and historical power load data based on support vector machine to build a power load prediction model;

[0083] Model evaluation unit: used to evaluate the performance of the constructed power load forecasting model and make necessary adjustments and optimizations;

[0084] The result prediction module further comprises:

[0085] Data collection window: used to collect regional image information of the target area according to the preset period;

[0086] Feature data extraction unit: used to extract power load related feature data from the collected regional portrait information;

[0087] Prediction execution unit: used to predict the extracted power load related characteristic data using the power load prediction model to generate the power load prediction result of the target area;

[0088] Result display unit: used to display the prediction results, including graphical display and report generation, to facilitate users to conduct further analysis and decision-making.

[0089] In this technical solution, the historical data integration unit: can effectively integrate historical power load data and power load-related feature data, provide comprehensive historical data support for model training, and improve the accuracy and reliability of the model. Feature alignment unit: ensure the consistency of power load-related features and historical data in the time dimension, avoid data mismatch problems, and improve the effectiveness of training data. Model training unit: supervised training based on support vector machine can effectively handle complex prediction tasks and improve the prediction accuracy of the power load prediction model. Model evaluation unit: evaluate the performance of the constructed prediction model, and make necessary adjustments and optimizations to ensure the efficiency and accuracy of the model. Data acquisition window: can regularly collect regional portrait information of the target area according to the preset cycle to ensure the timeliness and accuracy of the data. Feature data extraction unit: extract power load-related feature data from the collected regional portrait information, provide accurate data input for prediction, and improve the reliability of the prediction results. Prediction execution unit: use the constructed power load prediction model to predict the extracted feature data, generate the power load prediction results of the target area, and improve the pertinence and practicality of the prediction. Result display unit: provides graphical display and report generation to facilitate users to understand and analyze the prediction results and support further decision-making. This technical solution ensures the high accuracy and stability of the power load forecasting model through data integration, feature alignment, model training and evaluation. Through model evaluation and adjustment functions, the forecasting model can be continuously optimized to adapt to changes in the environment and demand and maintain the accuracy of the forecast results. The result display and analysis tools provided help users to deeply understand the power load forecasting results, thereby supporting the optimal scheduling and resource allocation of the power system.

[0090] Beneficial effects: The power load forecasting method based on regional portrait comprehensively collects relevant data of the target area, covering multiple dimensions such as economy, population, and industry. These data are then preprocessed to eliminate noise and redundant information to form an accurate and reliable regional benchmark data. After that, key indicators that can deeply reflect regional characteristics are extracted from the benchmark data using data mining technology. These indicators not only reflect the economic development status and industrial structure characteristics of the region, but also cover important factors such as population distribution and its changes. Based on these key indicators, a detailed and accurate target area portrait is constructed, which provides a strong basis for subsequent power load forecasting. After obtaining the regional portrait, features closely related to power load can be extracted more specifically. Then, these power load-related features and historical data are combined to construct an efficient and accurate power load forecasting model. Finally, the latest portrait information of the target area is collected in real time and input into the forecasting model to obtain the power load forecasting result of the target area. This method fully considers the characteristic differences of different regions, improves the accuracy of prediction, and provides strong support for the planning and management of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] Figure 1 It is a schematic diagram of the process of the present invention.

[0092] Figure 2 It is a system architecture diagram of the present invention.

[0093] In the figure: 1. Data processing module, 2. Data mining module, 3. Feature extraction module, 4. Model building module, 5. Result prediction module. DETAILED DESCRIPTION

[0094] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings.

[0095] Embodiment 1:

[0096] like Figure 1 As shown, this embodiment provides a method for predicting power load based on regional portraits, the method comprising:

[0097] S1: Collect regional relevant data of the target area, and pre-process the regional relevant data to obtain regional benchmark data;

[0098] Regional portraits are used to conduct comprehensive and in-depth analysis and description of specific geographical areas to reveal the characteristics, attributes and development trends of the region. Based on the collection, integration and analysis of a large amount of data, it covers multiple aspects such as economy, society and population, and aims to build a multi-dimensional and three-dimensional regional image, so that the system terminal can more accurately grasp the characteristics of different regions and extract key features related to power load, thereby building a more realistic prediction model and realizing personalized predictions for different regions.

[0099] In the embodiment of the present application, the system terminal collects regional relevant data of the target area, which is the first step in building a regional portrait and conducting power load forecasting. These data cover multiple aspects such as economy, society, population, geography, climate, energy consumption, industrial structure, etc., and are the basis for subsequent analysis and prediction. After collecting the data, the system terminal performs preprocessing, which includes cleaning, sorting, and standardization of the data to eliminate noise, correct errors, and ensure the accuracy and consistency of the data. After preprocessing, the system terminal defines the data as regional benchmark data, which will become an important reference for subsequent analysis and prediction.

[0100] Furthermore, the present application provides for collecting regional relevant data of the target area, and preprocessing the regional relevant data to obtain regional benchmark data, and the method also includes:

[0101] S101: Based on big data, collect regional relevant data of the target area;

[0102] S102: Perform data cleaning, normalization and missing value processing on the regional related data to obtain regional benchmark data, which includes regional data in multiple fields.

[0103] Preferably, the system terminal collects relevant data extensively for the target area based on big data technology. These data cover multiple fields such as economy, society, and population, providing a comprehensive and in-depth perspective for the system terminal. After that, these are processed. Among them, the data cleaning link can remove duplicate, erroneous or irrelevant information to ensure the purity of the data; normalization processing is to adjust the data to the same scale to facilitate subsequent analysis and comparison; missing value processing solves the vacancy problem in the data set, making the data more complete and reliable. After this series of processing, the system terminal obtains a comprehensive, accurate and reliable regional benchmark data, which will provide strong support for subsequent power load forecasting.

[0104] S2: Based on regional benchmark data, extract regional key indicators that reflect regional characteristics through data mining, and build a target regional profile based on regional key indicators;

[0105] In one embodiment, the system terminal further mines out the required indicator types based on the regional benchmark data that has been obtained, and analyzes and extracts the regional benchmark data according to the determined indicator types to obtain the required multiple regional key indicators. Using data mining can help the system terminal find key indicators that can truly reflect regional characteristics from a large amount of data. These key indicators cover multiple aspects such as economic development, social environment, population distribution, geographical location, climatic conditions, energy consumption, industrial structure, etc., and they are all important components of regional characteristics. After the system terminal extracts these key indicators, it analyzes trends and correlations based on these indicators to construct a portrait of the target area. This portrait not only shows the static characteristics of the region, such as geographical location characteristics, but also reflects the development trend and potential advantages of the region. By constructing a regional portrait, the characteristics and needs of the region can be understood more intuitively and deeply, providing a strong basis for subsequent power load forecasting.

[0106] Furthermore, the present application provides a method for extracting regional key indicators reflecting regional characteristics through data mining based on regional benchmark data, and the method also includes:

[0107] S201: Obtain key indicator types based on the power transmission structure of the target area;

[0108] S202: extracting indicators by principal component analysis and combining key indicator types to obtain multiple regional key indicators.

[0109] Preferably, when the system terminal predicts the power load based on the power transmission structure of the target area, it determines the key indicator types closely related to the structure. These key indicator types include grid scale, transmission capacity, line length, voltage level, power supply reliability, etc., which directly reflect the power transmission capacity and efficiency of the target area. For example, if the power transmission structure consists of power plants, transmission lines, substations and distribution networks, then corresponding indicators can be extracted for these four structures. By extracting the number of power plants, the number and type of substations, the grid scale indicator can be obtained; by extracting the total transmission capacity of the transmission line and the transmission capacity of the transmission lines of each voltage level, the transmission capacity indicator can be obtained. In order to extract the most representative indicators with the greatest impact on power load prediction from these indicator types, the system terminal also performs principal component analysis on these indicators. First, the covariance matrix of the regional benchmark data is calculated, which reflects the correlation between the indicators. Then, the eigenvalues ​​and eigenvectors of the covariance matrix are solved by mathematical methods. These eigenvectors represent the main direction of change of the data. Afterwards, the number of principal components is determined according to the size of the eigenvalues, and the explanatory power of the selected principal components for the regional benchmark data is evaluated to ensure that the selected principal components can fully represent the key features of the regional benchmark data. Then, the regional benchmark data is projected onto the selected principal components to obtain the coordinates of each sample in the new principal component space. These new coordinates are the regional key indicators extracted in combination with the key indicator types. These indicators obtained not only reflect the main characteristics of the power transmission structure of the target area, but also provide an important reference for subsequent power load forecasting.

[0110] Furthermore, the present application provides a method for constructing a target region profile based on regional key indicators, and the method further includes:

[0111] S203: Collect regional key indicator data, perform historical change trend analysis and indicator correlation analysis on the regional key indicators, and obtain change trend analysis results and indicator correlation analysis results;

[0112] Preferably, in order to deeply study the power load characteristics of the target area, the system terminal collects key indicator data closely related to the power transmission structure. These data include information on grid scale data, transmission capacity data, line length data, voltage level data, and other aspects. Once the data collection is completed, the system terminal performs historical change trend analysis on the key indicators of these regions. First, the collected key indicator data is preprocessed. This includes steps such as data cleaning, denoising, and standardization to eliminate the impact of outliers and unreasonable data on the analysis results, so that these data are more suitable for subsequent analysis. Afterwards, the indicator data in the past period of time in the processed key indicator data are statistically compared to analyze the laws and trends of each indicator over time. For example, in the data of the power grid of a large industrial city in the past five years, the grid scale data of the power grid showed a significant year-on-year expansion trend. The number of power plants has increased from 80 five years ago to 120 at present, and the number of substations has also increased from 150 to 220. This growth not only reflects the rapid increase in urban power demand, but also illustrates the steady progress of power grid infrastructure construction. The total transmission capacity has increased from 5,000 MW five years ago to 8,000 MW now, and the transmission capacity of transmission lines of various voltage levels has also increased. This shows that the transmission capacity of the power grid is constantly increasing and can better meet the growing power demand of cities. The total length of the lines is also increasing year by year, and the proportion of the length of high-voltage transmission lines is increasing year by year, which reflects that the power grid structure is developing in a more efficient and economical direction. Such historical change trend analysis helps the system terminal understand the historical evolution of the power transmission structure in the target area and provide an important reference for future predictions. At the same time, the system terminal sets the confidence level and significance level according to the data distribution, and calculates the correlation coefficient between each indicator. According to the calculated correlation coefficient, the correlation degree between the indicators is judged, and the indicator correlation analysis results are generated. When the correlation coefficient is closer to 1, the correlation degree between the two indicators is higher, and when it is close to 0, the correlation degree between the two indicators is lower. Through the above indicator correlation analysis, the system terminal can understand the mutual influence and correlation degree between different key indicators. For example, whether the expansion of the power grid will lead to an increase in transmission capacity, whether changes in line length will affect power supply reliability, etc. Through correlation analysis, the system terminal can reveal the intrinsic connection between these indicators and further understand the complexity and systematic nature of the power transmission structure in the target area. In summary, through historical trend analysis and indicator correlation analysis, the trend analysis results and indicator correlation analysis results can be obtained. These results not only help to fully understand the power transmission structure in the target area, but also provide important basis and support for the subsequent construction of the target area portrait.

[0113] S204: Based on the change trend analysis results and the indicator correlation analysis results, a basic framework of the target area portrait is constructed, including determining the main dimensions and sub-dimensions of the portrait;

[0114] Preferably, the main dimension refers to the key indicators that can comprehensively and systematically reflect the power transmission structure and development status of the target area, including the scale and capacity of power transmission, line construction and operation status, etc. The sub-dimension refers to the supplement and expansion of the main dimension, which is used to more deeply describe and characterize the detailed characteristics and conditions of the target area in specific aspects, including the coverage area of ​​the power grid, the length of each voltage level line, the proportion of different types of lines, the aging rate of the line, etc. The system terminal first sorts out the results of the change trend analysis and the indicator correlation analysis. This includes the historical change trend, change speed, fluctuation of key indicators, etc., as well as the correlation, dependency and influence mechanism between indicators. After that, according to the analysis results, the main dimensions of the target area portrait are determined. These dimensions can comprehensively reflect the core characteristics of the target area in terms of power transmission structure and development status. Under each main dimension, the system terminal further refines the sub-dimension. The sub-dimension is a supplement and expansion of the main dimension, which is used to more deeply describe and characterize the detailed characteristics and conditions of the target area in specific aspects. For example, under the dimension of power transmission scale and capacity, sub-dimensions such as grid coverage area, line length of each voltage level, and total transmission capacity are refined; under the dimension of line construction and operation status, sub-dimensions such as the proportion of different types of lines, line aging rate, and failure rate are refined. After determining the main dimensions and sub-dimensions, the system terminal integrates these dimensions to construct the basic framework of the target area portrait. This framework is a hierarchical structure, with the main dimensions as the top layer and the sub-dimensions as the lower layer, forming a clear and orderly structure. In summary, constructing the basic framework of the target area portrait based on the results of change trend analysis and indicator correlation analysis is a systematic and comprehensive process. By determining the main dimensions and sub-dimensions, the power portrait of the target area can be more accurately portrayed, providing strong support for the development of the power industry.

[0115] S205: According to the basic framework, the regional key indicator data is classified and placed to obtain a target regional portrait.

[0116] Preferably, according to the constructed basic framework of the target area portrait, the system terminal classifies and places the key indicator data of the region to obtain a complete target area portrait. This process is a systematic and logical work, which aims to integrate a large amount of indicator data into a unified framework, so as to intuitively show the comprehensive situation of the target area in terms of power transmission structure and development status. First, the system terminal classifies the indicator data according to the main dimensions. Each main dimension represents an important aspect, such as power transmission scale, line construction status, etc. The system terminal classifies the indicator data related to these dimensions into the corresponding dimensions to ensure the accuracy and completeness of the data. After that, the sub-dimensions are further refined under each main dimension, and the indicator data are placed in the corresponding sub-dimensions. The sub-dimensions provide a specific description and supplement to the main dimensions, making the portrait more detailed and in-depth. Through this step, the detailed characteristics and status of the target area in specific aspects can be more comprehensively understood. After completing the classification and placement, the system terminal obtains a target area portrait with a clear structure and rich content. This portrait not only shows the overall situation of the target area in terms of power transmission structure and development status, but also reveals its development trends, influencing factors and potential problems. In summary, by classifying and placing regional key indicator data according to the basic framework, a comprehensive, specific and in-depth portrait of the target region is obtained. This portrait provides a clear understanding of the power transmission structure and development status of the target region, which helps system terminals better understand and respond to power problems in the region.

[0117] S3: Extract power load related features based on the target area portrait;

[0118] In one embodiment, based on the constructed target area portrait, the system terminal extracts features related to the power load. These features reflect important information about the target area in terms of power demand, consumption patterns, and load changes, which are of great significance for power planning, scheduling, and management. The system terminal first extracts the power dispatching data of the target area and analyzes these data. Afterwards, the analysis results are combined with the constructed target area portrait to extract the power load-related features. In summary, based on the target area portrait, a wealth of power load-related features can be extracted. These features not only help the system terminal to fully understand the power demand status of the target area, but also provide strong data support for power planning. By analyzing and applying these features, the power demand of the target area can be better met, and the stable and safe operation of the power system can be guaranteed.

[0119] Furthermore, the present application provides a method for extracting power load related features based on a target area portrait, and the method further includes:

[0120] S301: Identify factors affecting power load according to power dispatching data of the target area;

[0121] S302: Based on the results of the power load influencing factors and indicator correlation analysis, the power load related features are extracted through the target area portrait.

[0122] Optionally, the system terminal identifies a variety of factors that affect the power load based on the power dispatch data of the target area. These factors include weather conditions, economic development, population growth trends, energy policy adjustments, etc. For example, in coastal cities, in hot summer weather, the use of refrigeration equipment such as air conditioners increases significantly, causing the power load to reach a peak. In severe weather conditions such as typhoons and rainstorms, power facilities may be damaged, resulting in power outages or load reduction. With the construction of industrial parks and commercial areas, population mobility has increased, resulting in a more significant growth rate of power load. Through in-depth analysis of power dispatch data, we can understand how these factors affect the fluctuations and changes in power demand. At the same time, combined with the previously obtained indicator correlation analysis results, we can further understand the interaction between each factor and the power load. This correlation analysis helps to clarify which factors are key influencers, which factors have mutual influence, and their specific degree of influence on the power load. With these influencing factors and correlation analysis results, the system terminal extracts features related to the power load through the target area portrait. The target area portrait provides a comprehensive and systematic perspective for the system terminal, enabling it to integrate various information to characterize the characteristics and laws of the power load. Through the portrait, specific features such as load peak and valley periods, load growth rate, and load volatility are extracted. These features can reflect the real-time status and development trend of the power load in the target area. In summary, through the combination of power dispatch data analysis and indicator correlation research, the influencing factors of the power load can be identified, and relevant features can be extracted through the target area portrait. This process not only helps the system terminal to deeply understand the formation mechanism of the power load, but also provides support for subsequent predictions.

[0123] S4: Based on the relevant characteristics of power load and combined with historical power load data, a power load forecasting model is constructed;

[0124] In one embodiment, the system terminal constructs a power load prediction model based on the extracted power load related features and combined with historical power load data. The main purpose of this model is to predict the power load in the future period of time, so as to provide a decision basis for the planning and scheduling of the power system. First, the system terminal uses the historical power load data and combines the previously extracted power load related features to perform data preprocessing and alignment operations. This step is to remove noise and irrelevant factors, ensure the accuracy of the data, and improve the prediction accuracy of the model. After that, a support vector machine is used to model according to the characteristics of the power load and the characteristics of the data, and the model is trained and optimized using historical data. By continuously adjusting the parameters and structure of the model, the model can better fit the historical data and have better generalization ability. In summary, the power load prediction model constructed by combining the power load related features and the historical power load data can accurately predict the future power load based on historical data and current features, and provide strong support for the planning and scheduling of the power system.

[0125] Furthermore, the present application provides a method for constructing a power load prediction model based on power load-related characteristics and in combination with historical power load data, and the method further includes:

[0126] S401: Collect historical power load data;

[0127] S402: Aligning power load related features with historical power load data based on a time dimension;

[0128] S403: Based on the support vector machine, supervised training is performed in combination with the power load related features and historical power load data to obtain a power load prediction model.

[0129] Preferably, in order to build a power load forecasting model, the system terminal first collects historical power load data, and checks and processes abnormal values ​​and missing values ​​in the data to ensure the integrity and accuracy of the data. These data record the changes in power load over a period of time in the past and are an important basis for predicting future power loads.

[0130] After the data is acquired, the system terminal compares the time range of the power load-related features and the historical power load data to ensure that they cover the same time period. If the time range is inconsistent, the system terminal cuts or fills the data according to the actual situation to unify the time range. Then, the power load-related feature data and the historical power load data are sorted according to the timestamps to ensure that the time order of the data is consistent. Then, the timestamps of the power load-related feature data and the historical power load data are matched one by one. If the timestamps match completely, the corresponding data is output directly; if they do not match completely, the closest time point is used for approximate matching. This step is to ensure that the feature data at each time point corresponds to the corresponding power load data, so that the features can accurately reflect the changes in the historical power load.

[0131] After completing the data alignment, the system terminal divides the aligned data set into a training set and a test set. The training set is used to train the power load forecasting model, while the test set is used to evaluate the forecasting performance of the power load forecasting model. Then, according to actual needs, select the corresponding kernel function and set the corresponding parameters to build the power load forecasting model. For example, when the power load data presents a linear or approximately linear relationship, select the linear kernel function and set the penalty coefficient to build the model so that the model can intuitively show the linear relationship between the features and the target. Afterwards, the aligned power load-related features and historical power load data in the training set are used to train the power load forecasting model.

[0132] During the training process, the system terminal sets the time window length, which is an adjustable parameter. The system terminal intercepts the training data by the time window length to obtain multiple first training data. Then, the system terminal performs constraint analysis on the obtained multiple first training data to determine whether these data have the characteristics of continuity and stability in this interval. If there are abnormal values ​​or mutation values, smoothing is performed, and multiple adjacent data are used for mean calculation to replace these abnormal values ​​or mutation values. Further, the system terminal inputs the multiple first training data after smoothing into the power load forecasting model to train the model. The power load forecasting model will learn the mapping relationship between power load-related features and power load data, and find the hyperplane that can maximize the classification interval. After that, the performance of the trained power load forecasting model is evaluated using the data in the test set, and the results of the model output are compared with the expected standards. If the expected standards are not met, the system terminal reselects the time window length, divides the new training data, and simulates different patterns that may appear in different time periods, and repeats the training process. If the expected effect is still not met after multiple iterations, the system terminal reselects the kernel function, configures the corresponding parameters, and repeats the above data division and training process until the expected effect is met. On the contrary, if the expected standard is met, the trained model is directly output to obtain the power load prediction model. This model can predict the power load situation in the future based on the input power load related characteristics. By continuously using new data to update and optimize the model, the prediction performance and accuracy of the model can be further improved.

[0133] S5: Collect regional portrait information of the target area, combine it with the power load prediction model to perform prediction, and obtain the power load prediction result of the target area.

[0134] In one embodiment, in order for the system terminal to obtain the power load forecast results of the target area, it is first necessary to collect the portrait information of the area. This information is the key basis for predicting the power load because they directly affect the pattern and scale of power demand. Afterwards, these regional portrait information are extracted and combined with the previously constructed power load forecasting model. The power load forecasting model has been trained with historical data and features, and has learned to infer future power loads from given features. In this process, the model predicts the power demand of the target area based on the information in the regional portrait. Finally, the model outputs the power load forecast results of the target area. This result is an estimate of the power load in the future, which can help power system operators to make power dispatch and planning in advance, ensure the stable operation of the power system and meet the power needs of users.

[0135] Query the actual power generation of the target PV power station and set the safe power generation threshold of the target PV power station;

[0136] Furthermore, the present application provides collecting regional portrait information of the target area, combining it with the power load prediction model to perform prediction, and obtaining the power load prediction result of the target area. The method also includes:

[0137] S501: Collecting regional portrait information of the target area through the data collection window according to a preset period;

[0138] S502: extracting power load-related characteristic data based on the regional portrait information;

[0139] S503: Use the power load related characteristic data in combination with the power load prediction model to perform prediction and obtain the power load prediction result.

[0140] Preferably, the system terminal collects regional portrait information of the target area according to a preset period through a set data collection window. After collecting this information, the system terminal further extracts feature data directly related to the power load using the same method as described above. Then, these extracted power load-related feature data are input into the previously trained power load forecasting model. Through the calculation of the model, the power load forecast result of the target area is obtained. In summary, by regularly collecting regional portrait information of the target area, extracting the power load-related feature data therein, and combining the power load forecasting model for prediction, the power load forecast result of the target area can be accurately obtained, providing strong data support for the planning, scheduling and operation of the power system.

[0141] For example, there is a city that decides to use a power load forecasting method based on regional portraits to improve the accuracy and reliability of power load forecasting. The city has four regions, A, B, C, and D, each with unique regional characteristics and power demand. The following is a detailed implementation process of the power load forecasting method based on regional portraits:

[0142] Collect regional data related to the target area, including population density, economic activities, electricity consumption, weather data, etc. The data format is JSON format:

[0143] {"A":{"Population density":1000,"Economic activity":5000,"Electricity consumption":300,"Weather":"Sunny"},

[0144] "B":{"Population density":1200,"Economic activity":4500,"Electricity consumption":320,"Weather":"Cloudy"},

[0145] "C":{"Population density":1100,"Economic activity":4800,"Electricity consumption":310,"Weather":"Rainy day"},

[0146] "D":{"Population density":1300,"Economic activity":5200,"Electricity consumption":330,"Weather":"Sunny"}}.

[0147] Perform data cleaning, normalization and missing value processing on regional related data to obtain regional benchmark data:

[0148] Region, population density, economic activity, electricity consumption, weather

[0149] A,1000,5000,300,sunny day

[0150] B,1200,4500,320,Cloudy

[0151] C,1100,4800,310,Rainy day

[0152] D,1300,5200,330,Sunny.

[0153] Based on the power transmission structure of the target area, key indicator types are obtained, such as power transmission path, power loss, etc. The principal component analysis method is combined with key indicator types to extract indicators and obtain multiple regional key indicators:

[0154] {"A":{"Key indicator 1":0.8,"Key indicator 2":0.6},

[0155] "B":{"Key indicator 1":0.7,"Key indicator 2":0.7},

[0156] "C":{"Key indicator 1":0.9,"Key indicator 2":0.5},

[0157] "D":{"Key indicator 1":0.6,"Key indicator 2":0.8}}.

[0158] Collect regional key indicator data, conduct historical change trend analysis and indicator correlation analysis on regional key indicators, and obtain change trend analysis results and indicator correlation analysis results:

[0159] {"A":{"Trend":"Increasing","Correlation":"High"},

[0160] "B":{"Trend":"Stable","Correlation":"Medium"},

[0161] "C":{"Trend":"Downward","Correlation":"Low"},

[0162] "D":{"Trend":"Increasing","Correlation":"High"}}.

[0163] Based on the results of the trend analysis and the indicator correlation analysis, the basic framework of the target area portrait is constructed, including determining the main dimensions and sub-dimensions of the portrait:

[0164] {"Main dimensions":["Population density","Economic activity","Electricity consumption","Weather"],

[0165] "Sub-dimension": ["Key Indicator 1", "Key Indicator 2"]}.

[0166] According to the basic framework, the regional key indicator data is classified and placed to obtain the target area portrait:

[0167] {"A":{"Population density":1000,"Economic activity":5000,"Electricity consumption":300,"Weather":"Sunny","Key indicator 1":0.8,"Key indicator 2":0.6},

[0168] "B":{"Population density":1200,"Economic activity":4500,"Electricity consumption":320,"Weather":"Cloudy","Key indicator 1":0.7,"Key indicator 2":0.7},

[0169] "C":{"Population density":1100,"Economic activity":4800,"Electricity consumption":310,"Weather":"Rainy day","Key indicator 1":0.9,"Key indicator 2":0.5},

[0170] "D":{"Population density":1300,"Economic activity":5200,"Electricity consumption":330,"Weather":"Sunny","Key indicator 1":0.6,"Key indicator 2":0.8}}.

[0171] Identify the factors affecting power load based on the power dispatch data of the target area:

[0172] {"A":{"Influencing Factor 1":0.7,"Influencing Factor 2":0.5},

[0173] "B":{"Influencing Factor 1":0.6,"Influencing Factor 2":0.6},

[0174] "C":{"Influencing Factor 1":0.8,"Influencing Factor 2":0.4},

[0175] "D":{"Influencing factor 1":0.5,"Influencing factor 2":0.7}}.

[0176] Based on the analysis results of the influencing factors and index correlation of power load, the relevant features of power load are extracted through the target area portrait:

[0177] {"A":{"Load characteristic 1":0.75,"Load characteristic 2":0.55},

[0178] "B":{"Load characteristic 1":0.65,"Load characteristic 2":0.65},

[0179] "C":{"Load characteristic 1":0.85,"Load characteristic 2":0.45},

[0180] "D":{"Load characteristic 1":0.55,"Load characteristic 2":0.75}}.

[0181] Collect historical power load data:

[0182] Date, power consumption

[0183] 2024-05-01,280

[0184] 2024-05-02,290

[0185] 2024-05-03,300.

[0186] Based on the time dimension, the power load related characteristics are aligned with the historical power load data:

[0187] Date, Load Characteristics 1, Load Characteristics 2, Electricity Consumption

[0188] 2024-05-01,0.75,0.55,280

[0189] 2024-05-02,0.65,0.65,290

[0190] 2024-05-03,0.85,0.45,300.

[0191] Based on the support vector machine, combined with its power load related features and historical power load data, supervised training is carried out to obtain the power load forecasting model:

[0192] {"Model":"Support Vector Machine",

[0193] "Parameters":{"C":1.0,"gamma":0.1}}.

[0194] Through the data collection window, collect regional image information of the target area according to the preset cycle:

[0195] {"A":{"Population density":1050,"Economic activity":5100,"Electricity consumption":310,"Weather":"Sunny"},

[0196] "B":{"Population density":1250,"Economic activity":4600,"Electricity consumption":330,"Weather":"Cloudy"},

[0197] "C":{"Population density":1150,"Economic activity":4900,"Electricity consumption":320,"Weather":"Rainy day"},

[0198] "D":{"Population density":1350,"Economic activity":5300,"Electricity consumption":340,"Weather":"Sunny"}}.

[0199] Based on the regional profile information, the power load related characteristic data is extracted to obtain the power load forecast results:

[0200] {"A":{"Load characteristic 1":0.76,"Load characteristic 2":0.56},

[0201] "B":{"Load characteristic 1":0.66,"Load characteristic 2":0.66},

[0202] "C":{"Load characteristic 1":0.86,"Load characteristic 2":0.46},

[0203] "D":{"Load characteristic 1":0.56,"Load characteristic 2":0.76}}.

[0204] In summary, the embodiments of the present application have at least the following technical effects:

[0205] The embodiment of the present application collects and preprocesses the regional relevant data of the target area to obtain the regional benchmark data. Based on the regional benchmark data, the key indicators reflecting the regional characteristics are extracted through data mining. This includes determining the key indicator type based on the power transmission structure and using the principal component analysis method to extract the indicators. Afterwards, these regional key indicators are used to construct a target area portrait. Through historical change trend analysis and indicator correlation analysis, the main dimensions and sub-dimensions of the portrait are determined, and the key indicator data are classified and placed. Subsequently, the power load-related features are extracted according to the target area portrait. Then, a power load forecasting model is constructed in combination with historical power load data. Finally, by regularly collecting the regional portrait information of the target area and combining it with the forecasting model for forecasting, the power load forecasting results are obtained. This provides strong data support for the planning, scheduling and operation of the power system. These technical effects jointly solve the technical problem that the traditional power load forecasting method is difficult to adapt to the differences in characteristics of different regions, and improve the practicality and pertinence of the forecasting results.

[0206] Embodiment 2

[0207] Based on the same inventive concept as the power load forecasting method based on regional portrait in the aforementioned embodiment, Figure 2 As shown, the present application provides a power load forecasting system based on regional portrait, the system comprising:

[0208] Data processing module 1: Data processing module 1 is used to collect regional relevant data of the target area, and pre-process the regional relevant data to obtain regional benchmark data;

[0209] Data mining module 2: Data mining module 2 is used to extract regional key indicators reflecting regional characteristics through data mining based on regional benchmark data, and to construct a target regional profile based on the regional key indicators;

[0210] Feature extraction module 3: Feature extraction module 3 is used to extract power load related features based on the target area portrait;

[0211] Model building module 4: Model building module 4 is used to build a power load prediction model based on power load related characteristics and historical power load data;

[0212] Result prediction module 5: Result prediction module 5 is used to collect regional portrait information of the target area, and perform prediction in combination with the power load prediction model to obtain the power load prediction result of the target area.

[0213] Furthermore, the data processing module 1 is used to execute the following method:

[0214] Based on big data, collect regional relevant data of the target area;

[0215] The regional related data are cleaned, normalized and missing values ​​processed to obtain regional benchmark data, which includes regional data in multiple fields.

[0216] Furthermore, the data mining module 2 is used to perform the following method:

[0217] Based on the power transmission structure of the target area, obtain key indicator types;

[0218] Through principal component analysis and combining key indicator types, indicators are extracted to obtain multiple regional key indicators.

[0219] Furthermore, the data mining module 2 is used to perform the following method:

[0220] Collect regional key indicator data, conduct historical change trend analysis and indicator correlation analysis on regional key indicators, and obtain change trend analysis results and indicator correlation analysis results;

[0221] Based on the results of trend analysis and indicator correlation analysis, the basic framework of the target area portrait is constructed, including determining the main dimensions and sub-dimensions of the portrait;

[0222] According to the basic framework, regional key indicator data are classified and placed to obtain a portrait of the target area.

[0223] Furthermore, the data mining module 2 is used to perform the following method:

[0224] Identify factors affecting power load based on power dispatch data in the target area;

[0225] Based on the results of the correlation analysis of factors affecting power load and indicators, the power load-related characteristics are extracted through the target area portrait.

[0226] Furthermore, the model building module 4 is used to perform the following method:

[0227] Collect historical power load data;

[0228] Based on the time dimension, the power load related features are aligned with the historical power load data;

[0229] Based on support vector machine, combined with its power load related features and historical power load data, supervised training is carried out to obtain the power load forecasting model.

[0230] Furthermore, the result prediction module 5 is used to perform the following method:

[0231] Through the data collection window, regional image information of the target area is collected according to the preset period;

[0232] Extract power load-related characteristic data based on regional profile information;

[0233] The power load prediction results are obtained by using the power load related characteristic data combined with the power load prediction model.

[0234] The power load forecasting method and system based on regional portrait shown above are specific embodiments of the present invention, which have embodied the essential characteristics and progress of the present invention. They can be equivalently modified according to actual usage needs under the guidance of the present invention, and are all within the protection scope of this scheme.

Claims

1. A method for power load forecasting based on regional portrait, characterized in that Includes steps: 1) Collecting regional relevant data of the target area and preprocessing the regional relevant data to obtain regional benchmark data; 2) Based on the regional benchmark data, extract regional key indicators reflecting regional characteristics through data mining, and use the regional key indicators to construct a target regional profile; 3) Extracting power load related features based on the target area portrait; 4) Building a power load forecasting model based on the power load related characteristics and combining historical power load data; 5) Collecting regional portrait information of the target area, and performing prediction in combination with the power load prediction model to obtain the power load prediction result of the target area; In step 2), based on the regional benchmark data, regional key indicators reflecting regional characteristics are extracted through data mining, including: 201) Determine the key indicator types based on the power transmission structure of the target area; the key indicator types include grid scale, transmission capacity, line length, voltage level and power supply reliability; 202) extracting indicators by principal component analysis in combination with the key indicator types to obtain multiple regional key indicators; Calculate the covariance matrix of regional benchmark data, which reflects the correlation between various indicators; Use mathematical methods to solve the eigenvalues ​​and eigenvectors of the covariance matrix. The eigenvector represents the main direction of change of the data. According to the size of the eigenvalues, the number of principal components is determined, and the explanatory power of the selected principal components on the regional benchmark data is evaluated to ensure that the selected principal components can fully represent the key features of the regional benchmark data; Project the regional benchmark data onto the selected principal component to obtain the coordinates of each sample in the new principal component space, which are the regional key indicators extracted in combination with the key indicator type; In step 2), a target region profile is constructed based on the regional key indicators, including: 203) Collect regional key indicator data, conduct historical change trend analysis and indicator correlation analysis on the regional key indicators, and obtain change trend analysis results and indicator correlation analysis results; the key indicator data include power grid scale data, transmission capacity data, line length data and voltage level data; 204) Based on the above trend analysis results and indicator correlation analysis results, a basic framework for the target area portrait is constructed, including determining the main dimensions and sub-dimensions of the portrait; the main dimensions refer to key indicators that can comprehensively and systematically reflect the power transmission structure and development status of the target area, including power transmission scale and capacity, line construction and operation status; the sub-dimensions refer to the supplement and expansion of the main dimensions, which are used to more deeply describe and characterize the detailed characteristics and status of the target area in specific aspects, including the grid coverage area, the length of lines of each voltage level, the proportion of different types of lines and the line aging rate; the basic framework is a hierarchical structure, with the main dimensions as the top layer and the sub-dimensions as the bottom layer, so as to form a clear and orderly structure; 205) According to the basic framework, the key indicator data of the region are classified and organized to obtain a target region portrait, which includes the static characteristics, development trends and potential advantages of the region; Step 3) includes: 301) Based on the power dispatch data of the target area, identify the factors affecting power load; the influencing factors include weather conditions, economic development, population growth trends and energy policy adjustments; 302) Based on the power load influencing factors and indicator correlation analysis results, power load related features are extracted through the target area portrait.

2. The method for power load forecasting based on regional portrait according to claim 1, characterized in that: Step 1) includes: 101) Based on big data, collect regional relevant data of the target area; 102) Performing data cleaning, normalization processing and missing value processing on the regional related data to obtain the regional benchmark data, wherein the regional benchmark data includes regional data in multiple fields.

3. The method for power load forecasting based on regional profile according to claim 1, characterized in that: Step 4) includes: 401) Collect historical power load data; 402) Based on the time dimension, aligning the power load related features with the historical power load data to ensure that the time periods covered by the two are the same; If the time ranges of the two are inconsistent, the data is trimmed or filled according to the actual situation to unify the time range; the power load-related characteristic data and the historical power load data are sorted according to the timestamps to ensure that the time sequence of the data is consistent; then, the timestamps of the power load-related characteristic data and the historical power load data are matched one by one; If the timestamps match exactly, the corresponding data is output directly; If there is not an exact match, use the closest time point for an approximate match; 403) Based on a support vector machine, supervised training is performed on the power load related features and historical power load data to construct a power load prediction model; The aligned data set is divided into a training set and a test set; the training set is used to train the power load forecasting model, and the test set is used to evaluate the prediction performance of the power load forecasting model; Then, according to actual needs, select the corresponding kernel function and set the corresponding parameters to build the power load forecasting model; when the power load data presents a linear or approximately linear relationship, select the linear kernel function and set the penalty coefficient to build the model so that the model can intuitively show the linear relationship between the feature and the target; Using the aligned power load related features and historical power load data in the training set, training the power load forecasting model; During the training process, a time window length is set, and the training data is intercepted by the time window length to obtain a plurality of first training data; a constraint analysis is performed on the obtained plurality of first training data to determine whether these data have the characteristics of continuity and stability in this interval; if there are abnormal values ​​or mutation values, smoothing is performed, and a plurality of adjacent data are used to perform mean calculation to replace these abnormal values ​​or mutation values; the plurality of first training data after smoothing is input into the power load prediction model to train the model; the power load prediction model learns the mapping relationship between the power load related features and the power load data, and finds a hyperplane that can maximize the classification interval; Use the data in the test set to evaluate the performance of the trained power load forecasting model, and compare the model output results with the expected standards. If the expected standards are not met, reselect the time window length, divide new training data, simulate different patterns that may appear in different time periods, and repeat the training process. If the expected effect is still not met after multiple iterations, the kernel function is reselected, the corresponding parameters are configured, and the above data division and training process is repeated until the expected effect is met; conversely, if the expected standard is met, the trained model is directly output to obtain the power load prediction model.

4. The method for power load forecasting based on regional profile according to claim 1, characterized in that: Step 5) includes: 501) Collecting regional image information of the target area according to a preset period through the data collection window; 502) Extracting power load related characteristic data based on the regional portrait information; 503) Using the power load related characteristic data and the power load prediction model to perform prediction, and obtain a power load prediction result.

5. A power load forecasting system using the power load forecasting method based on regional profile according to any one of claims 1 to 4, characterized in that include: Data processing module: used to collect regional relevant data of the target area and pre-process these data to obtain regional benchmark data; Data mining module: Based on regional benchmark data, extract regional key indicators that reflect regional characteristics through data mining, and use these key indicators to build a target regional profile; Feature extraction module: extracts power load related features based on the target area portrait; Model building module: constructs a power load forecasting model based on power load-related characteristics and historical power load data; Result prediction module: collects regional portrait information of the target area, and combines it with the power load prediction model to make predictions to obtain the power load prediction results of the target area.

6. The power load forecasting system according to claim 5, characterized in that: The data processing module further comprises: Data collection unit: used to connect to the data source of the target area through an interface and collect area-related data in real time or periodically; Data preprocessing unit: used to clean, normalize and fill missing values ​​of the collected regional related data to generate regional benchmark data; Data storage unit: used to store pre-processed regional benchmark data and support subsequent data mining and analysis operations; The data mining module further comprises: Key indicator type determination unit: determines the key indicator type according to the power transmission structure of the target area; Principal component analysis unit: extracting indicators by principal component analysis and combining the key indicator types to obtain multiple regional key indicators; Data analysis unit: used to collect regional key indicator data, and perform historical change trend analysis and indicator correlation analysis on the regional key indicators, to obtain change trend analysis results and indicator correlation analysis results; Portrait construction unit: Based on the results of change trend analysis and indicator correlation analysis, the basic framework of the target area portrait is constructed, including determining the main dimensions and sub-dimensions of the portrait; Data organization unit: classify and organize regional key indicator data according to the basic framework to obtain a target regional portrait; The feature extraction module further comprises: Power load feature identification unit: used to identify key factors affecting power load based on power dispatch data of the target area; Feature extraction unit: used to extract power load related feature data based on power load influencing factors and regional portrait information; Feature storage unit: used to store the extracted power load-related feature data for further analysis and training by the model building module.

7. The power load forecasting system according to claim 6, characterized in that: The model building module further comprises: Historical data integration unit: used to collect and integrate historical power load data and power load related characteristic data; Feature alignment unit: used to align power load related features with historical power load data based on the time dimension to ensure data consistency; Model training unit: used to conduct supervised training on power load related features and historical power load data based on support vector machine to build a power load prediction model; Model evaluation unit: used to evaluate the performance of the constructed power load forecasting model and make necessary adjustments and optimizations; The result prediction module further comprises: Data collection window: used to collect regional image information of the target area according to the preset period; Feature data extraction unit: used to extract power load related feature data from the collected regional portrait information; Prediction execution unit: used to predict the extracted power load related characteristic data using the power load prediction model to generate the power load prediction result of the target area; Result display unit: used to display the prediction results, including graphical display and report generation, to facilitate users to conduct further analysis and decision-making.

Images