Short-term Electric Load Forecasting in Microgrids: Issues and Challenges

Abstract

This paper compares performance of three well-known short-term load forecasting (STLF) methodologies in microgrid applications. The chosen methods include: i) seasonal auto-regressive integrated moving average with exogenous variables, ii) neural networks, and iii) wavelet neural networks. These methods utilise combinations of historical load data and metrological variables to predict the load of individual customers in a microgrid over the next day. This is essential for scheduling, management and control of microgrid resources. So far, the existing STLF methodologies have been successfully used for the aggregated load forecasting in transmission and distribution systems. Nevertheless, their prediction accuracy in microgrid applications, where diversity is low and considerable changes in the load of customers can be observed in a short period of time, is not investigated. The random and chaotic nature of individual customers’ loads make STLF challenging; hence, this paper aims to address the issues for the above methodologies in microgrids.