Power load forecasting by neural network with a new learning process for considering overtraining problem

Abstract

In this paper, a neural network (NN) with a new learning process is proposed for power load forecasting to overcome the problem of over-training. This new learning process is developed to solve the problems of underfitting, resulting from under-training, and overfitting, resulting from over-training. As a comparison of the traditional method of cross-validation (CV) and our proposed learning process, Taipower load signals and relevant weather information from 1990 to 1993 are investigated.