Ultra-short-term power load prediction method and device based on machine learning

Abstract

The invention discloses an ultra-short-term power load prediction method based on machine learning. The method comprises the following steps: carrying out normalization operation on original data; processing load data by using a variational mode decomposition method, and decomposing original load data with relatively strong volatility into a plurality of relatively stable intrinsic mode function components; clustering each intrinsic mode function component by using a Kmeans method to obtain a new load sequence; performing feature construction by considering the influence of influence factors on the power load; and predicting the new load sequence by using an XGBoost model, and then superposing and reconstructing predicted values to obtain a final prediction result. According to the method, the influence of factors such as weather, holidays and the like on data feature construction is considered, and the advantages of VMD, Kmeans and XGBoost methods are combined, so that the ultra-short-term load prediction precision can be improved.

Description

technical field [0001] The invention relates to a machine learning-based ultra-short-term power load forecasting method and device, and belongs to the technical field of power data processing. Background technique [0002] Power load forecasting is one of the technical prerequisites to ensure the safe and stable operation of power systems, and ultra-short-term load forecasting is of great significance for real-time security analysis and real-time economic dispatch. With the rapid development of the current economy, the demand for electricity in various industries is getting higher and higher, and the load types are becoming more and more complex. The power load is affected by factors such as temperature, humidity, and holidays, and its randomness is strong. Therefore, improving the accuracy of load forecasting is a current research main goal of . [0003] Research on load forecasting models at home and abroad is mainly divided into time series models, machine learning and d...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problem that the power load is affected by the complex environment with strong randomness and low prediction accuracy, this invention considers the influence of weather and holidays to construct data features and establishes a model through data mining and machine learning algorithms, and proposes a machine learning-based ultra-short-term power Load forecasting method and device

Images