Adjustment of discrete load changes in feeder databases for improving medium-term demand forecasting

Abstract

A discrete load change (DLC) event may be described as an abrupt change in feeder demand. These events are due to network reconfigurations, or the connection/disconnection of large consumers to the grid. This phenomenon affects the performance of load forecasting methods and, in general, it may worsen any planning or operational application that uses feeder demand records as input. This study proposes four load adjustment (LA) methods to correct this type of distortion from distribution system demand database. The methods are tested by using real demand values, encompassing six years of hourly data registered in 169 feeders, of a distribution company. To test the effectiveness of the LA methods, in medium-term load forecasting, a comparative study using different forecasting techniques is performed. Results show that demand forecasting, with DLC adjustment, improve their average performance over 33% compared to the case were this phenomena is not considered.