One-day-ahead load forecast using an adaptive approach

Abstract

Electrical load forecasting is vitally important to modern power system planning, operation, and control. In this paper, by focusing on historical load data and calendar factors, we present a hybrid method using period refinement scheme and adaptive strategy for building peak hour period and off-peak hour period models in day-of-week for one-day-ahead for load forecasting. They are evaluated using three full years of Shenzhen city electricity load data. Experimental results shows the adaptive model for each period, confirm good accuracy of the proposed approach to load forecasting and indicate that it has better forecasting accuracy than traditional ANN method.