Load forecasting method based on experience day local correction multipoint extrapolation method

The invention provides a load forecasting method based on an experience day local correction multipoint extrapolation method. The method includes the following steps that (1) sample data are selected, wherein the date type, the time before the forecasting day and the like are mainly considered, and no more than two months before/after the forecasting day, or several historical days of the same month of the previous year of the forecasting day are taken to form a sample set of load forecasting; (2) abnormal data are detected and processed, wherein data vertical processing, data horizontal processing and missing data processing are included; (3) the change tendency of a further load is forecasted through probability statistics, the load change tendency of two adjacent points is divided into a rising tendency, a falling tendency and a level tendency; and the further load change tendency is forecasted through a principle of probability statistics analysis; (4) forecasting is conducted on the points to be forecasted through the local correction multipoint extrapolation method one by one; the future load change value is estimated by the adoption of the least square method, and the method for improving the accuracy of a load forecasting model mainly is the multipoint extrapolation method by the adoption of experience day local correction.