Power System Load Forecasting Method Based on Multivariate Data Fusion

Abstract

The present invention discloses a power system load forecasting method based on multi-source data fusion, which relates to the technical field of power systems. In order to solve the problems in the prior art that the multi-source data fusion processing is complex, its timeliness and processing efficiency are insufficient, and the response ability to short-term emergencies is weak under complex working conditions; the present invention can comprehensively consider various factors affecting load changes by collecting and preprocessing multi-source data in real time, enhances the response ability to complex working conditions and emergencies, dynamically adjusts the time series window to improve data validity and prediction performance, improves the accuracy and reliability of power system load forecasting, constructs and trains multiple models to obtain accurate load forecasting results, realizes the intelligent forecasting and management of power system load through accurate load forecasting, and combines with the emergency event database for anomaly identification and processing, provides stable and reliable load forecasting for the power system, thereby optimizing the power resource allocation.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and particularly to a power system load forecasting method based on multi-source data fusion. Background Art

[0002] Traditional load forecasting methods often rely solely on historical load data, ignoring other factors that may affect load changes, such as weather conditions, socioeconomic status, user behavior patterns, etc. Chinese Patent with the publication number CN118780449A discloses a power load forecasting method, which includes acquiring power system data; preprocessing and multi-source data fusion processing of the power system data; using a variational mode decomposition model to decompose the original time series into subsequences; constructing a power load forecasting model based on a gated recurrent unit and an attention mechanism, and using the power load forecasting model to forecast the power load according to the decomposed subsequences, wherein an improved sparrow search algorithm is used to optimize the parameters of the power load forecasting model. This patent solves the problems of insufficient accuracy of traditional power load forecasting models, difficulty in optimizing model parameters, and poor multi-source data fusion processing effects.

[0003] However, although the above patent improves the forecasting accuracy to a certain extent, there are still the following problems:

[0004] The multi-source data fusion processing in the prior art is complex, with insufficient timeliness and processing efficiency, and has a weak response ability to short-term emergencies under complex working conditions. Summary of the Invention

[0005] The purpose of the present invention is to provide a power system load forecasting method based on multi-source data fusion. By integrating and analyzing information from multiple data sources, it can comprehensively consider various factors affecting load changes, enhance the response ability to complex working conditions and emergencies, improve the accuracy and reliability of power system load forecasting, and achieve intelligent forecasting and management of power system loads through accurate load forecasting, so as to solve the problems raised in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solutions:

[0007] A power system load forecasting method based on multi-source data fusion includes the following steps:

[0008] Step 1: Data preprocessing: Real-time multi-source data is collected based on multi-source acquisition terminals. The multi-source data specifically includes historical load data, meteorological data, holiday information, and economic index data. The obtained multi-source data is subjected to data cleaning and normalization processing, and feature encoding processing is performed based on the multi-source data categories.

[0009] Step 2: Multi-source data fusion: Perform fusion and feature optimization on the preprocessed multi-source data, screen out the key features highly relevant to load forecasting, and dynamically adjust the time series window according to the short-term and long-term load forecasting requirements;

[0010] Step 3: Multi-model load forecasting: Build corresponding load forecasting models based on the multi-source data categories and train them. Use the trained load forecasting models to perform load forecasting to obtain the forecasting results of future loads;

[0011] Step 4: Forecast result evaluation: Evaluate the model accuracy through model evaluation metrics. Based on the evaluation results, adjust the structure and parameters of the load forecasting model in combination with the real-time forecasting accuracy feedback data, and visually display the final future load forecasting results;

[0012] Step 5: Abnormal response: Establish an emergency event database, identify the abnormal types of the forecasting results based on the emergency event database, and perform abnormal handling responses according to the abnormal types.

[0013] Further, the real-time collection of multi-source data by the multi-source collection terminal specifically includes:

[0014] Data preprocessing: Perform deduplication and removal of irrelevant information on the collected multi-source data, and perform standardized conversion on the multi-source data with different dimensions;

[0015] Feature encoding processing: Perform one-hot encoding and label encoding on the categorical features of the multi-source data. Among them, one-hot encoding is used to convert the categorical features of holiday information and meteorological data into binary vectors, and label encoding is used to process the categorical variables of historical load data;

[0016] Consistency check: Align the timestamps and sampling frequencies carried by the multi-source data, check and unify the measurement units and scales of the multi-source data, and generate continuous and synchronized time series data.

[0017] Further, Step 2: Multi-source data fusion is specifically as follows:

[0018] Multi-source data fusion: Perform feature-level fusion on the multi-source data after consistency check to construct a unified feature matrix;

[0019] Feature recognition: Calculate the correlation coefficient between each feature in the feature matrix and the electric load, and identify the key features related to the electric load;

[0020] Feature screening: Calculate the importance weights of each key feature, and recursively eliminate the key features with the lowest weights until the preset number of features is reached to determine the final subset of key features;

[0021] Dynamic adjustment of time series window: Dynamically adjust the size of the time window according to the load prediction requirements of different time granularities, determine the change characteristics of the load data under different time lengths, and optimize the window length according to the feedback data of the prediction accuracy using the sliding window method.

[0022] Furthermore, the dynamic adjustment of the time series window includes:

[0023] Extract the change rate corresponding to the current load data;

[0024] Compare the change rate corresponding to the load data with a preset change rate threshold;

[0025] When the change rate corresponding to the load data exceeds the preset change rate threshold, retrieve the time granularity corresponding to the current time series window;

[0026] Take the duration corresponding to the time granularity of the current time series window as the unit reference time;

[0027] Extract the load data in each unit reference time;

[0028] Use the load data in each unit reference time to obtain the load coefficient corresponding to each unit reference time;

[0029] Among them, the load coefficient corresponding to each unit reference time is obtained through the following formula:

[0030] ;

[0031] Among them, U represents the load coefficient corresponding to each unit reference time; n represents the number of load data acquisitions corresponding to each unit reference time; P represents the change rate value corresponding to when the preset change rate threshold is exceeded; W i represents the load data value corresponding to the i-th load data acquisition; W i+1 represents the load data value corresponding to the (i + 1)-th load data acquisition; W z represents the intermediate value of the load data corresponding to n load data acquisitions;

[0032] Retrieve the intermediate value of the load coefficient from the load coefficients corresponding to each unit reference time;

[0033] Compare the intermediate value of the load coefficient with a preset load coefficient threshold;

[0034] When the intermediate value of the load coefficient exceeds the preset load coefficient threshold, dynamically adjust the time series window.

[0035] Furthermore, the dynamic adjustment of the time series window includes:

[0036] When the intermediate value of the load factor exceeds the preset load factor threshold, retrieve the maximum and minimum time granularities corresponding to the historical load data;

[0037] Use the duration corresponding to the maximum time granularity as the first unit reference time;

[0038] Use the duration corresponding to the minimum time granularity as the second unit reference time;

[0039] Under the first unit reference time, obtain the load factor corresponding to each first unit reference time as the first load factor dataset;

[0040] Under the second unit reference time, obtain the load factor corresponding to each second unit reference time as the second load factor dataset;

[0041] Use the load factors included in the first load factor dataset and the second load factor dataset, combined with the intermediate value of the load factor corresponding to the current unit reference time, to obtain the time window adjustment coefficient;

[0042] Among them, the time window adjustment coefficient is obtained through the following formula:

[0043] ;

[0044] Among them, K represents the time window adjustment coefficient; m represents the number of load factors included in the first load factor dataset; k represents the number of load factors included in the second load factor dataset; U z represents the intermediate value of the load factor corresponding to the current unit reference time; U 01i represents the corresponding value of the i-th load factor in the first load factor dataset; U 02i represents the corresponding value of the i-th load factor in the second load factor dataset; U 01z represents the intermediate value of the coefficients corresponding to the m load factors in the first load factor dataset; U 02z represents the intermediate value of the coefficients corresponding to the k load factors in the second load factor dataset; T 01 and T 02 represent the durations corresponding to the first unit reference time and the second unit reference time respectively;

[0045] Use the time window adjustment coefficient to adjust the time granularity of the time window to obtain the adjusted time granularity;

[0046] Among them, the duration corresponding to the adjusted time granularity is obtained through the following formula:

[0047] ;

[0048] Among them, Tt Indicates the corresponding duration of the adjusted time granularity; T 0 Indicates the corresponding duration of the time granularity before adjustment; K represents the time window adjustment coefficient.

[0049] Furthermore, the feature recognition further includes:

[0050] Obtain power data from the power system, perform redundancy and dimensionality reduction processing on the power data, extract features from the processed power data, and obtain the initial feature set of the power data according to the extraction results;

[0051] Identify and extract key features from the initial feature set. At the same time, use each key feature in the key feature subset as a model output sample to train a preset network model to generate a load recognition model;

[0052] Based on the load recognition model, obtain the target power consumption characteristics corresponding to the target time series data of each key feature in the key feature subset;

[0053] Obtain the change situation of the target power consumption characteristics of each load in the power data, and determine the power consumption change rule of each load according to the change situation.

[0054] Furthermore, the prediction result of the future load obtained in step three includes:

[0055] Obtain the prediction accuracy of each key feature as the first index for evaluating the model performance;

[0056] Calculate the dynamic change characteristic value of each key feature as the second index for model performance evaluation;

[0057] Count the frequency and data coverage rate of each key feature in the data set as supplementary performance indicators;

[0058] According to the prediction accuracy, dynamic characteristic value and data coverage rate of each key feature, calculate the confidence level of the corresponding feature, and convert the confidence level into a weight value;

[0059] Based on the weight value, perform a weighted average calculation on the prediction results of the power data corresponding to each data label to obtain the final load prediction result, and quantify the importance of each key feature in the prediction result;

[0060] Based on the weight values of each key feature, perform a weighted average process on the prediction results of each key feature output by the load prediction model, and generate the final future load prediction result based on the weighted calculation result.

[0061] Furthermore, the step of generating the final future load prediction result in step three further includes determining the change mode of the power load and identifying the power consumption law. The specific steps are as follows:

[0062] Analysis of electricity consumption change patterns:

[0063] Based on the electricity consumption change rules for each load amount, identify and classify the electricity consumption characteristics during peak, trough, and flat peak periods, and analyze the fluctuation patterns of the load at different times;

[0064] Combine meteorological data to determine the change rules of the load affected by meteorological factors, and identify the formation reasons for seasonal peaks and troughs;

[0065] Based on holiday information, analyze the load fluctuation characteristics during holidays, distinguish the electricity consumption rules between normal working days and holidays, and identify the load change patterns during special periods;

[0066] Analysis of economic activity correlations:

[0067] Based on economic indicator data, evaluate the correlation with the electricity consumption change rules of the load amount, establish a correlation model between the load change and economic indicator data, and predict the long-term growth trend of the load;

[0068] Comprehensive correlation analysis:

[0069] Based on the fused multi-dimensional data, construct a multi-dimensional correlation model for power load changes, and obtain the coupling relationship between the load change and multi-dimensional data;

[0070] Based on the trained load prediction model, verify the identified electricity consumption change rules, and optimize the electricity consumption pattern by combining the multi-source data obtained in real time.

[0071] Furthermore, the model evaluation indicators in step four specifically include: mean absolute error indicator, mean square error indicator, root mean square error indicator, mean absolute percentage error indicator.

[0072] Furthermore, the abnormal types of the identified prediction results in step five specifically include:

[0073] Classify and mark the emergency information in the emergency event database, and associate the mark with the time stamp, affected area, and load change data of the corresponding emergency event information;

[0074] Compare the residuals between the actual load data and the prediction results, identify the abnormal data points beyond the normal fluctuation range, match the detected abnormal data points with the emergency event information, and identify the abnormal types based on the matching results;

[0075] Re-evaluate the prediction results after abnormal processing, identify the load change patterns before and after the event occurs, and verify the abnormal processing effect;

[0076] Establish a feedback mechanism between anomaly handling and load forecasting results, and adjust the load forecasting model based on real-time prediction accuracy feedback data.

[0077] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0078] By collecting and preprocessing multi-source data in real time, various factors affecting load changes can be comprehensively considered, enhancing the response ability to complex working conditions and emergencies, dynamically adjusting the time series window to improve data validity and prediction performance, improving the accuracy and reliability of power system load forecasting, constructing and training multiple models to obtain accurate load forecasting results, using evaluation indicators such as mean absolute error and mean square error to ensure prediction accuracy, achieving intelligent forecasting and management of power system load through accurate load forecasting, and combining with an emergency event database for anomaly identification and handling, providing stable and reliable load forecasting for the power system, thereby optimizing the allocation of power resources and ensuring the safe and economic operation of the system. Description of the Drawings

[0079] Figure 1 It is a flowchart of the power system load forecasting method based on multi-source data fusion of the present invention. Detailed Embodiments

[0080] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0081] To solve the technical problems in the prior art that the multi-source data fusion processing is complex, its timeliness and processing efficiency are insufficient, and the response ability to short-term emergencies is weak under complex working conditions, please refer to Figure 1 , the following technical solutions are provided in this embodiment,

[0082] A power system load forecasting method based on multi-source data fusion includes the following steps:

[0083] Step 1: Data preprocessing: Based on multi-source acquisition terminals, historical load data, meteorological data, holiday information, economic indicators and other multi-source data are collected in real time, and the obtained multi-source data is subjected to data cleaning and normalization processing, missing values and outliers are processed, and feature encoding processing is performed based on the multi-source data categories to ensure the consistency and integrity of the multi-source data;

[0084] Step 2: Multi-source data fusion: Fuse and optimize the features of the preprocessed multi-source data, screen out the key features highly relevant to load forecasting, and dynamically adjust the time series window according to the short-term and long-term load forecasting requirements to improve the effectiveness and forecasting performance of the multi-source data;

[0085] Step 3: Multi-model load forecasting: Build corresponding load forecasting models based on the multi-source data categories and train them. Divide the preprocessed multi-source data into a training set and a test set, train the corresponding load forecasting models based on the training set data, and use the trained load forecasting models to perform load forecasting to obtain the forecasting results of future loads;

[0086] Among them, the load forecasting models include time series models, regression models, neural network models, and ensemble learning models, etc. The training process includes processing such as data segmentation, loss function optimization, and gradient descent, and adjusting the structure and parameters of the model through real-time feedback data.

[0087] In this embodiment, the time series model processes the time series data of power loads and can capture the seasonal and trend changes of loads, such as ARIMA, SARIMA, LSTM, etc. In particular, LSTM has strong advantages in capturing long-term dependencies and dynamic change trends;

[0088] In this embodiment, the regression model is used to establish the relationship between load data and multi-source features, such as linear regression, support vector regression SVR, XGBoost, etc.;

[0089] In this embodiment, the neural network model automatically extracts complex features from multi-source data through a deep learning model to optimize the forecasting performance, such as multi-layer perceptron, convolutional neural network CNN, etc.;

[0090] In this embodiment, the ensemble learning model can improve the accuracy and stability of forecasting, and is suitable for the fusion of multiple features and complex pattern recognition, such as random forest, AdaBoost, etc.;

[0091] In this embodiment, multiple models such as long short-term memory network (LSTM), support vector regression (SVR), and random forest are used for load forecasting, so as to be able to process multi-source heterogeneous data and select and adjust the model according to the actual data characteristics and forecasting accuracy.

[0092] Step 4: Forecast result evaluation: Evaluate the model accuracy through model evaluation metrics. Based on the evaluation results, adjust the structure and parameters of the load forecasting model in combination with the real-time forecasting accuracy feedback data to ensure the accuracy and reliability of the forecasting results, and visually display the final future load forecasting results;

[0093] The model evaluation metrics specifically include: the mean absolute error metric, which is used to measure the average deviation between the predicted value and the actual value; the mean squared error metric, which is used to evaluate the average of the squares of the differences between the predicted value and the actual value; the root mean squared error metric, which represents the magnitude of the prediction error and has the same dimension as the actual value; the mean absolute percentage error metric, which is used to measure the prediction accuracy;

[0094] Step Five: Abnormal Response: Establish an emergency event database to record events that may cause abnormal load, such as weather disasters, equipment failures, etc. Based on the emergency event database, identify the abnormal types of the prediction results and perform abnormal handling responses according to the abnormal types.

[0095] In this embodiment, by real-time collecting and preprocessing multi-source data, various factors affecting load changes can be comprehensively considered, enhancing the response ability to complex working conditions and emergency events, dynamically adjusting the time series window to improve data validity and prediction performance, improving the accuracy and reliability of power system load prediction, constructing and training multiple models to obtain accurate load prediction results, using evaluation metrics such as mean absolute error and mean squared error to ensure prediction accuracy, through accurate load prediction, realizing intelligent prediction and management of the power system load, and combining the emergency event database for abnormal identification and processing, providing stable and reliable load prediction for the power system, thereby optimizing the allocation of power resources and ensuring the safe and economic operation of the system.

[0096] In this embodiment, the real-time collection of multi-source data based on the multi-source collection terminal specifically includes:

[0097] Determine the data source: Real-time collect the historical load data of the power grid through the power system load monitoring terminal, and obtain meteorological data such as temperature, humidity, wind speed, and rainfall based on means such as meteorological stations and satellite remote sensing, obtain holiday information, and collect economic indicator data;

[0098] Data preprocessing: Perform deduplication and removal of irrelevant information operations on the collected multi-source data to ensure data quality, and perform standardized conversion on multi-source data with different dimensions for subsequent analysis;

[0099] Feature encoding processing: Perform one-hot encoding and label encoding processing on the categorical features of multi-source data. Among them, one-hot encoding is used to convert the categorical features of holiday information and meteorological data into binary vectors, and label encoding is used to process the categorical variables of historical load data, decompose the timestamps of historical load data, and extract time features, such as peak hours, mid-peak hours, and low-peak hours;

[0100] Consistency check: Align the timestamps and sampling frequencies carried by multi-source data to ensure the consistency of multi-source data in the time dimension. Check and unify the measurement units and scales of multi-source data to ensure that the data has the same measurement standard, and generate continuous and synchronized time series data, which is processed using linear interpolation or forward filling methods.

[0101] In this embodiment, by accurately determining the data sources and collecting multi-source data in real time, performing efficient data preprocessing and quality assurance, feature encoding processing and time feature extraction, as well as consistency check and generation of time series data, the deep fusion and efficient utilization of multi-source data are achieved, significantly improving the accuracy and timeliness of power system load forecasting. It not only ensures the consistency and integrity of the data, but also enhances the model's ability to identify complex load change patterns through refined feature processing, thereby effectively improving the accuracy of load forecasting, providing reliable data support and technical guarantee for the stable operation of the power system and resource optimization configuration, and having obvious practical application benefits.

[0102] In this embodiment, Step 2: Multi-source data fusion, specifically:

[0103] Multi-source data fusion: Perform feature-level fusion on the multi-source data after consistency check to construct a unified feature matrix. The feature matrix includes all relevant features to ensure that all features are consistent under a unified time scale, such as the topological structure information of the distribution network, node attributes, time series features, etc.

[0104] Feature recognition: Calculate the correlation coefficient between each feature in the feature matrix and the power load, and identify the key features related to the power load.

[0105] Feature screening: Calculate the importance weights of each key feature, quantify the contribution degree of each feature to the prediction result, and recursively remove the key features with the lowest weights until the preset number of features is reached to determine the final subset of key features.

[0106] Dynamic adjustment of time series window: Dynamically adjust the size of the time window according to the load forecasting requirements of different time granularities (such as hours, days, weeks), determine the change characteristics of the load data under different time lengths, and use the sliding window method to optimize the window length according to the prediction accuracy feedback data to adapt to short-term and long-term load forecasting requirements, ensuring a high degree of matching between the time series features and the load changes.

[0107] Specifically, the dynamic adjustment of the time series window includes:

[0108] Extract the change rate corresponding to the current load data;

[0109] Compare the change rate corresponding to the load data with a preset change rate threshold;

[0110] When the change rate corresponding to the load data exceeds a preset change rate threshold, the time granularity corresponding to the current time series window is retrieved;

[0111] Take the duration corresponding to the time granularity of the current time series window as the unit reference time;

[0112] Extract the load data in each unit reference time;

[0113] Use the load data in each unit reference time to obtain the load coefficient corresponding to each unit reference time;

[0114] Among them, the load coefficient corresponding to each unit reference time is obtained through the following formula:

[0115] ;

[0116] Among them, U represents the load coefficient corresponding to each unit reference time; n represents the number of load data acquisition times corresponding to each unit reference time; P represents the change rate value corresponding to when the preset change rate threshold is exceeded; W i represents the load data value corresponding to the i-th load data acquisition; W i+1 represents the load data value corresponding to the (i + 1)-th load data acquisition; W z represents the intermediate value of the load data corresponding to n times of load data acquisition;

[0117] Retrieve the intermediate value of the load coefficient from the load coefficients corresponding to each unit reference time;

[0118] Compare the intermediate value of the load coefficient with a preset load coefficient threshold;

[0119] When the intermediate value of the load coefficient exceeds the preset load coefficient threshold, the time series window is dynamically adjusted.

[0120] The technical effects of the above technical solution are as follows: By extracting the change rate of the current load data in real time and comparing it with a preset change rate threshold, this solution can dynamically identify the fluctuations in the load data. When the load data changes significantly (i.e., the change rate exceeds the threshold), the solution can respond quickly and dynamically adjust the time series window. This dynamic adjustment mechanism enables the solution to more flexibly adapt to different load change patterns, improving the adaptability and response speed of the system. The solution calculates the load coefficient within each unit reference time and uses this coefficient to evaluate the stability and change trend of the load data. The calculation of the load coefficient takes into account multiple factors such as the number of load data acquisitions, the change rate value, and the median value of the load data, so as to more accurately reflect the actual situation of the load data. This refined processing method helps to improve the analysis accuracy and accuracy of the system for load data. By using the time granularity of the current time series window as the unit reference time and extracting the load data within each unit reference time, the solution can more effectively utilize data resources and reduce the interference of redundant information. At the same time, by comparing the median value of the load coefficient with a preset load coefficient threshold, the time series windows that need to be adjusted can be further screened out, thus improving the processing efficiency and stability of the system. Parameters such as the change rate threshold and the load coefficient threshold in this solution can be set and adjusted according to actual needs, making the solution have a certain degree of scalability and customizability. This helps to meet the requirements in different application scenarios and improve the applicability and flexibility of the solution. By dynamically adjusting the time series window, this solution can more reasonably allocate system resources and avoid unnecessary calculations and processing when the load data fluctuates slightly. This helps to reduce the energy consumption and cost of the system and improve the resource utilization efficiency.

[0121] In summary, the technical effects of this technical solution in terms of performance indicators are mainly reflected in aspects such as dynamic adaptability, precision and accuracy, efficiency and stability, scalability and customizability, and optimizing resource allocation and reducing energy consumption. These effects together enhance the overall performance and user experience of the system.

[0122] Specifically, the dynamic adjustment of the time series window includes:

[0123] When the median value of the load coefficient exceeds the preset load coefficient threshold, retrieve the maximum time granularity and the minimum time granularity corresponding to the historical load data;

[0124] Take the duration corresponding to the maximum time granularity as the first unit reference time;

[0125] Take the duration corresponding to the minimum time granularity as the second unit reference time;

[0126] Under the first unit reference time, obtain the load coefficient corresponding to each first unit reference time as the first load coefficient data set;

[0127] At the second unit reference time, obtain the load factor corresponding to each second unit reference time as the second load factor dataset;

[0128] Use the load factors included in the first load factor dataset and the second load factor dataset, and combine them with the load factor median corresponding to the current unit reference time to obtain the time window adjustment coefficient;

[0129] Among them, the time window adjustment coefficient is obtained through the following formula:

[0130] ;

[0131] Among them, K represents the time window adjustment coefficient; m represents the number of load factors included in the first load factor dataset; k represents the number of load factors included in the second load factor dataset; U z represents the load factor median corresponding to the current unit reference time; U 01i represents the corresponding value of the i-th load factor in the first load factor dataset; U 02i represents the corresponding value of the i-th load factor in the second load factor dataset; U 01z represents the coefficient median value corresponding to the m load factors in the first load factor dataset; U 02z represents the coefficient median value corresponding to the k load factors in the second load factor dataset; T 01 and T 02 respectively represent the corresponding durations of the first unit reference time and the second unit reference time;

[0132] Use the time window adjustment coefficient to adjust the time granularity of the time window to obtain the adjusted time granularity;

[0133] Among them, the corresponding duration of the adjusted time granularity is obtained through the following formula:

[0134] ;

[0135] Among them, T t represents the corresponding duration of the adjusted time granularity; T 0 represents the corresponding duration of the time granularity before adjustment; K represents the time window adjustment coefficient.

[0136] The technical effects of the above technical solution are as follows: By introducing a time window adjustment coefficient, the above technical solution realizes the dynamic adjustment of the time granularity of the time window. When the median value of the load factor exceeds the preset threshold, the above technical solution can automatically retrieve the time granularity information of historical load data, and calculate the time window adjustment coefficient based on this information and the current median value of the load factor. This dynamic adjustment mechanism enables the solution to better adapt to the fluctuations of load data, improving the accuracy and efficiency of analysis. By obtaining the load factor data sets at the first unit reference time and the second unit reference time respectively, and combining with the current median value of the load factor, the solution can more comprehensively consider the changes of load data at different time granularities. This processing method helps to improve the accuracy and stability of load factor calculation, thus more accurately reflecting the actual characteristics of load data. By dynamically adjusting the time granularity of the time window, the above technical solution can reasonably allocate system resources according to the fluctuations of actual load data. When the load data fluctuates greatly, a smaller time granularity is adopted to capture data changes more precisely; when the load data is relatively stable, a larger time granularity is adopted to reduce the calculation amount. This optimization strategy helps to improve the utilization efficiency of system resources and reduce the calculation cost. At the same time, the time window adjustment coefficient calculation formula and the time granularity adjustment formula in the above technical solution both have certain scalability and flexibility. By adjusting the parameters and coefficients in the formula, the requirements in different application scenarios can be met, improving the applicability and flexibility of the solution. Through the dynamic adjustment of the time granularity of the time window, the solution can provide more accurate and timely load data analysis results, providing more powerful support for decision-making. This helps to improve the decision-making efficiency, reduce the decision-making risk, and improve the overall operation efficiency.

[0137] In summary, the technical effects of this technical solution in terms of performance indicators are mainly reflected in aspects such as dynamic adjustment and adaptability, improving accuracy and stability, optimizing resource utilization, scalability and flexibility, and enhancing decision-making efficiency. These effects jointly improve the overall performance and practical value of the solution.

[0138] In this embodiment, the feature recognition further includes:

[0139] Obtain power data from the power system, including real-time data such as the load, voltage, and current of each distribution node, perform redundancy and dimensionality reduction processing on the power data, and perform feature extraction on the processed power data. Obtain the initial feature set of the power data according to the extraction results, including load features, time series features, etc.;

[0140] Identify and extract key features from the initial feature set. At the same time, use each key feature in the key feature subset as a model output sample to train a preset network model to generate a load identification model;

[0141] Obtain the target power consumption characteristics corresponding to the target time series data of each key feature in the key feature subset based on the load quantity recognition model;

[0142] Obtain the change situation of the target power consumption characteristics of each load quantity in the power data, and determine the power consumption change rule of each load quantity according to the change situation.

[0143] In this embodiment, a unified feature matrix is constructed through feature-level fusion, combined with feature recognition, feature screening and dynamic adjustment of the time series window, effectively improving the accuracy and adaptability of load forecasting. Feature recognition not only calculates the correlation coefficient between features and power load, but also obtains real-time data from the power system to establish an initial feature set. This process further refines the recognition of key features. The load quantity recognition model obtains the target power consumption characteristics corresponding to the target time series data of key features, ensuring that the power consumption change rule of each load quantity can be accurately determined, thereby improving the model's ability to capture the change trend of power load, significantly enhancing the adaptability and prediction accuracy of the load forecasting model to complex power system conditions, and providing strong technical support for the stable operation and resource optimization of the power system.

[0144] In this embodiment, the prediction result of the future load obtained in step three includes:

[0145] Obtain the prediction accuracy of each key feature as the first index for evaluating the model performance;

[0146] Calculate the dynamic change characteristic values of each key feature (such as load fluctuation amplitude, trend stability) as the second index for model performance evaluation;

[0147] Count the frequency and data coverage of each feature in the data set as supplementary performance indicators;

[0148] According to the prediction accuracy, dynamic characteristic values and data coverage of each key feature, calculate the confidence of the corresponding feature and convert the confidence into a weight value;

[0149] Based on the weight value, perform a weighted average calculation on the prediction results of the power data corresponding to each data label to obtain the final load prediction result, quantifying the importance of each key feature in the prediction result;

[0150] Based on the weight values of each key feature, perform a weighted average process on the prediction results of each key feature output by the load forecasting model, and generate the final future load prediction result based on the weighted calculation result, improving the overall accuracy and reliability of the prediction result.

[0151] In this embodiment, by introducing prediction accuracy, dynamic characteristic values, and data coverage rate, a comprehensive model evaluation system is formed. Weights are dynamically allocated based on the importance of each key feature to enhance the model's adaptability to complex data, effectively integrate the prediction results of different features, and reduce the impact of single-feature errors on the overall prediction.

[0152] In this embodiment, the step three of generating the final future load prediction result further includes determining the change pattern of the power load and identifying the electricity consumption pattern. The specific steps are as follows:

[0153] Analysis of electricity consumption change pattern:

[0154] Based on the electricity consumption change rules of each load quantity, identify and classify the electricity consumption characteristics during peak, valley, and flat peak periods, and analyze the fluctuation patterns of the load at different times;

[0155] Combine meteorological data to determine the change rules of the load affected by meteorological factors, and identify the formation reasons of seasonal peaks and valleys;

[0156] Based on holiday information, analyze the load fluctuation characteristics during holidays, distinguish the electricity consumption patterns between normal working days and holidays, and identify the load change patterns during special periods;

[0157] Analysis of economic activity correlation:

[0158] Based on economic indicator data such as GDP, industrial added value, and resident consumption level, evaluate the correlation with the electricity consumption change rules of the load quantity, establish a correlation model between the load change and economic indicator data, and predict the long-term growth trend of the load;

[0159] Comprehensive correlation analysis:

[0160] Based on the fused multi-dimensional data such as time, season, and economic activities, construct a multi-dimensional correlation model of power load change, and obtain the coupling relationship between the load change and multi-dimensional data;

[0161] Based on the trained load prediction model, verify the identified electricity consumption change rules to ensure the accuracy and applicability of the rules, and optimize the electricity consumption pattern by combining the real-time obtained multi-source data.

[0162] In this embodiment, the electricity consumption change rules are integrated into the load prediction model to enhance the model's adaptability to complex scenarios, help the model more accurately capture the load change trend. By identifying the electricity consumption change rules of each load quantity, determine the change pattern of the power load over time, season, economic activities and other factors, identify the electricity consumption rules during peak and valley periods, and monitor the long-term change trend of the load with economic growth and industrial structure adjustment.

[0163] In this embodiment, identifying the abnormal types of the prediction results in step five specifically includes:

[0164] Classify and label the emergency event information in the emergency event database, and associate the label with the timestamp, affected area, and load change data of the corresponding emergency event information;

[0165] Compare the residuals between the actual load data and the prediction results, identify the abnormal data points beyond the normal fluctuation range, match the detected abnormal data points with the emergency event information, and identify the abnormal types based on the matching results;

[0166] The abnormal types are as follows: Meteorological anomaly: caused by extreme weather events, such as high temperature, snowstorm, etc.; Equipment failure anomaly: caused by power grid equipment failure, maintenance, or accidents; Economic activity anomaly: caused by sudden economic policies, sudden increase or decrease in industrial activities; Holiday anomaly: load change caused by holidays or major social events;

[0167] Re-evaluate the prediction results after anomaly processing to ensure that the corrected load prediction conforms to the actual situation, identify the load change patterns before and after the event, and verify the effect of anomaly processing;

[0168] Establish a feedback mechanism between anomaly processing and the load prediction results, and adjust the load prediction model based on the real-time prediction accuracy feedback data.

[0169] In this embodiment, through refined data preprocessing, in-depth identification and matching of abnormal types, and establishment of an effective feedback mechanism, the accuracy and robustness of load prediction are significantly improved. It can timely capture and accurately analyze the load change trend, effectively identify and respond to the impacts of various emergencies on the power system load, ensure the stability and reliability of power supply, and at the same time provide strong data support for the operation scheduling, resource optimization configuration, and risk management of the power system, thereby improving the overall operation efficiency and economy of the power system and meeting the demand for high-quality power supply services.

[0170] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A power system load forecasting method based on multivariate data fusion, characterized in that: The following steps are involved: Step 1: Data preprocessing: Multi-source data is collected in real time based on the multi-source collection terminal. The multi-source data specifically includes historical load data, meteorological data, holiday information, and economic indicator data. The acquired multi-source data is cleaned and normalized, and feature coding is performed based on the multi-source data category. Step 2: Multi-source data fusion: Fusion and feature optimization of pre-processed multi-source data, screening out key features that are highly relevant to load forecasting, and dynamically adjusting the timing window based on short-term and long-term load forecasting requirements; Dynamic adjustment of timing window, including: Extract the change rate corresponding to the current load data; Comparing the change rate corresponding to the load data with a preset change rate threshold; When the change rate corresponding to the load data exceeds a preset change rate threshold, the time granularity corresponding to the current time series window is retrieved; Taking the time granularity corresponding to the duration of the current timing window as the unit reference time; Extract load data in each unit reference time; Obtaining a load factor corresponding to each unit reference time by using the load data in each unit reference time; The load factor corresponding to each unit reference time is obtained by the following formula: ； Among them, U represents the load factor corresponding to each unit reference time; n represents the number of load data collections corresponding to each unit reference time; P represents the corresponding change rate value when exceeding the preset change rate threshold; W i represents the load data value corresponding to the i-th load data collection; W i+1 represents the load data value corresponding to the i+1th load data collection; W z Indicates the middle value of load data corresponding to n load data collections; Retrieve the middle value of the load factor from the load factor corresponding to each unit reference time; comparing the load factor intermediate value with a preset load factor threshold; When the load factor intermediate value exceeds a preset load factor threshold, the timing window is dynamically adjusted; Step 3: Multi-model load forecasting: Build and train the corresponding load forecasting model based on multi-source data categories, and use the trained load forecasting model to perform load forecasting to obtain the forecast results of future loads; Step 4: Forecast result evaluation: Evaluate the model accuracy through model evaluation indicators, adjust the structure and parameters of the load forecast model based on the evaluation results and combined with real-time forecast accuracy feedback data, and visualize the final future load forecast results; Step 5: Abnormal response: Establish an emergency event database, identify the abnormal type of the prediction result based on the emergency event database, and perform abnormal processing response according to the abnormal type.

2. The method for power system load forecasting based on multivariate data fusion according to claim 1, characterized in that: The real-time collection of multi-source data based on the multi-source collection terminal specifically includes: Data preprocessing: remove duplicates and irrelevant information from the collected multi-source data, and standardize and convert multi-source data of different dimensions; Feature encoding processing: Perform one-hot encoding and label encoding processing on the category features of multi-source data. Among them, one-hot encoding is used to convert the category features of holiday information and meteorological data into binary vectors, and label encoding is used to process the categorical variables of historical load data; Consistency check: Align the timestamps and sampling frequencies carried by multi-source data, check and unify the measurement units and scales of multi-source data, and generate continuous and synchronized time series data.

3. The method for power system load forecasting based on multivariate data fusion according to claim 2, characterized in that: Step 2: Multi-source data fusion, specifically: Multi-source data fusion: After consistency verification, multi-source data is fused at the feature level to build a unified feature matrix; Feature identification: Calculate the correlation coefficient between each feature in the feature matrix and the power load, and identify the key features related to the power load; Feature screening: Calculate the importance weight of each key feature and recursively remove the key features with the lowest weight until the preset number of features is reached to determine the final key feature subset; Dynamic adjustment of time window: Dynamically adjust the time window size according to the load forecasting requirements of different time granularities, determine the changing characteristics of load data at different time lengths, and use the sliding window method to optimize the window length based on the forecast accuracy feedback data.

4. The method for power system load forecasting based on multivariate data fusion according to claim 3, characterized in that: Dynamically adjust the timing window, including: When the load factor intermediate value exceeds the preset load factor threshold, the maximum time granularity value and the minimum time granularity value corresponding to the historical load data are retrieved; Taking the duration corresponding to the maximum value of the time granularity as the first unit reference time; Taking the duration corresponding to the minimum value of the time granularity as the second unit reference time; Under the first unit reference time, obtaining a load factor corresponding to each first unit reference time as a first load factor data set; Under the second unit reference time, obtaining a load factor corresponding to each second unit reference time as a second load factor data set; Obtaining a time window adjustment coefficient using the load factors included in the first load factor data set and the second load factor data set combined with an intermediate value of the load factors corresponding to the current unit reference time; The time window adjustment coefficient is obtained by the following formula: ； Wherein, K represents the time window adjustment coefficient; m represents the number of load factors included in the first load factor data set; k represents the number of load factors included in the second load factor data set; U z Indicates the middle value of the load factor corresponding to the current unit reference time; U 01i represents the value corresponding to the i-th load factor in the first load factor data set; U 02i represents the value corresponding to the i-th load factor in the second load factor data set; U 01z represents the middle value of the coefficient corresponding to the m load factors in the first load factor data set; U 02z represents the middle value of the coefficient corresponding to the k load factors in the second load factor data set; T 01 and T 02 Respectively represent the corresponding durations of the first unit reference time and the second unit reference time; Using the time window adjustment coefficient to adjust the time granularity of the time window, and obtain the adjusted time granularity; The corresponding duration of the adjusted time granularity is obtained by the following formula: ； Among them, T t Indicates the duration corresponding to the time granularity after adjustment; T0 indicates the duration corresponding to the time granularity before adjustment; K indicates the time window adjustment coefficient.

5. The method for power system load forecasting based on multivariate data fusion according to claim 3, characterized in that: The feature recognition further includes: Acquire power data from the power system, perform redundancy and dimensionality reduction processing on the power data, perform feature extraction on the processed power data, and obtain an initial feature set of the power data according to the extraction result; Identify and extract key features from the initial feature set, and use each key feature in the key feature subset as a model output sample to train a preset network model to generate a load recognition model; Obtaining target power consumption characteristics corresponding to target time series data of each key feature in the key feature subset based on the load identification model; The change of the target power consumption characteristic of each load in the power data is obtained, and the power consumption change rule of each load is determined according to the change.

6. The method for power system load forecasting based on multivariate data fusion according to claim 3, characterized in that: The prediction result of future load obtained in step 3 includes: Obtain the prediction accuracy of each key feature as the first indicator for evaluating model performance; Calculate the dynamic change characteristic value of each key feature as the second indicator for model performance evaluation; Count the frequency and data coverage of each key feature in the data set as a supplementary performance indicator; According to the prediction accuracy, dynamic characteristic value and data coverage of each key feature, the confidence of the corresponding feature is calculated and converted into a weight value; Based on the weight value, the prediction results of the power data corresponding to each data tag are weighted averaged to obtain the final load forecast result, and the importance of each key feature in the forecast result is quantified; Based on the weight value of each key feature, the prediction results of each key feature output by the load forecasting model are weighted averaged, and the final future load forecast result is generated based on the weighted calculation result.

7. The method for power system load forecasting based on multivariate data fusion according to claim 6, characterized in that: The step 3 generates the final future load forecast result, and also includes determining the change pattern of the power load and identifying the power consumption pattern. The specific steps are as follows: Analysis of electricity consumption change patterns: Based on the electricity consumption variation rules of each load, identify and classify the electricity consumption characteristics during peak, valley and off-peak periods, and analyze the load fluctuation patterns during different periods; Combined with meteorological data, determine the changing pattern of load affected by meteorological factors and identify the causes of seasonal peaks and troughs; Analyze the load fluctuation characteristics during holidays based on holiday information, distinguish the power consumption patterns during normal working days and holidays, and identify the load change patterns during special periods; Economic activity correlation analysis: Based on the correlation between economic indicator data evaluation and load power consumption change rules, a correlation model between load change and economic indicator data is established to predict the long-term growth trend of load; Comprehensive correlation analysis: Based on the fused multidimensional data, a multidimensional correlation model of power load changes is constructed to obtain the coupling relationship between load changes and multidimensional data; The identified electricity consumption variation patterns are verified based on the trained load forecasting model, and the electricity consumption pattern is optimized by combining multi-source data obtained in real time.

8. The method for power system load forecasting based on multivariate data fusion according to claim 7, characterized in that: The model evaluation indicators in step 4 specifically include: mean absolute error indicator, mean square error indicator, root mean square error indicator, and mean absolute percentage error indicator.

9. The method for power system load forecasting based on multivariate data fusion according to claim 8, characterized in that: The abnormal type of the prediction result identified in step 5 specifically includes: Classify and mark the emergency event information in the emergency event database, and associate the mark with the timestamp, impact area and load change data of the corresponding emergency event information; Compare the residuals between the actual load data and the predicted results, identify abnormal data points that exceed the normal fluctuation range, match the detected abnormal data points with the emergency event information, and identify the abnormal type based on the matching results; Re-evaluate the prediction results after exception handling, identify the change pattern of load before and after the event, and verify the effect of exception handling; A feedback mechanism between exception handling and load forecasting results is established, and real-time forecast accuracy feedback data is obtained based on the feedback mechanism.

Images