Method and evaluation method of ultra-short-load forecasting in power system

Abstract

This article describes the basic method of ultra-short-term load forecasting include the Linear Extrapolation, Kalman Filter Method, Time Series Method, Artificial Neural Networks and Support Vector Machine Algorithm. Then, it summarizes the commonly used methods to improve the accuracy of prediction from the comprehensive forecasting model and data mining technology. Finally, we divide the accuracy of the short-term load forecast into 5 min, 10 min, 30 min and 60 min and innovatively presented the concept of the accuracy in daily average load forecasting from different periods by giving the evaluation formula.