A genetic programming approach to forecast daily electricity demand

Abstract

A number of recent researches have compared machine learning techniques to find more reliable approaches to solve variety of engineering problems. In the present study, capability of canonical genetic programming (GP) technique to model daily electrical energy consumption (ED) as an alternative for electrical demand prediction was investigated. For this aim, using the most recent ED data recorded at northern part of Nicosia, Cyprus, we put forward two daily prediction scenarios subjected to train and validate by GPdotNET, an open source GP software. Minimizing root mean square error between the modeled and observed data as the objective function, the best prediction model at each scenario has been presented for the city. The results indicated the promising role of GP for daily ED prediction in Nicosia, however it suffers from lagged prediction that must be considered in practical application.