Short-term electrical energy demand prediction under heat island effects using emotional neural network integrated with genetic algorithm

Abstract Electricity, considered to be the lifeblood of industrialized societies, is underpinned by a vastly complex and interconnected grid consisting of multiple stakeholders. In this industry, the accurate forecasting of electricity load has become vital to planning, operations, transmission safety, energy transaction planning, and economic dispatch. However, accurate forecasting is a challenge due to a range of uncertainties with weather conditions and extreme weather events chief among them. The forecasting challenge is magnified by the existence of urban climate phenomena such as the Urban Heat Islands. These have the effect of intensifying weather conditions and changing electricity demand patterns, especially during summer months. Intending to address this challenge, we evaluate a hybrid model, the winner-take-all emotional neural network, against random forest and multiple linear regression models for the short-term electricity demand prediction utilizing air temperature data from fixed weather stations and reanalysis data from the European Centre for Medium-Range Weather Forecasts.