A BUS BAR LOAD FORECASTING METHOD

摘要

A bus bar load forecasting method, the method comprising: correcting abnormal values among historical load data using a lateral comparative method (202), and determining key influencing factors of bus bar load using a grey association projection method (203); classifying load curves with similar characteristics into one category using an improved K-means clustering method to obtain a plurality of typical load patterns (204); building a random forest classification model, and establishing a mapping relationship between influence factors and clustering results (205); for each category of load patterns, training a plurality of forecasting models using a multivariate linear regression method (206); and determining, by means of the random forest classification model, the category of a day to be tested, and selecting a matching regression model to realize load forecasting (207). The method introduces a data mining method to analyze the change rule of bus bar load and establish a forecasting model library, and realize model matching in combination with a day to be tested, improving the accuracy and real-time performance of short-term bus bar load forecasting, providing more accurate decision support for power grid planning and real-time scheduling.