Preprocessing data for short-term load forecasting with a general regression neural network and a moving average filter

Abstract

This paper proposes a filter based on a general regression neural network and a moving average filter, for preprocessing half-hourly load data for short-term multinodal load forecasting, discussed in another paper. Tests made with half-hourly load data from nine New Zealand electrical substations demonstrate that this filter is able to handle noise, missing data and abnormal data. © 2011 IEEE.