Prediction of electricity consumption for residential houses in New Zealand

Abstract

Residential consumer’s demand of electricity is continuously growing, which leads to high greenhouse gas emissions. Detailed analysis of electricity consumption characteristics for residential buildings is needed to improve efficiency, availability and plan in advance for periods of high electricity demand. In this research work, we have proposed an artificial neural network based model, which predicts the energy consumption of a residential house in Auckland 24 hours in advance with more accuracy than the benchmark persistence approach. The effects of five weather variables on energy consumption was analyzed. Further, the model was experimented with three different training algorithms, the levenberg-marquadt (LM), bayesian regularization and scaled conjugate gradient and their effect on prediction accuracy was analyzed.