Prediction of household electricity consumption and effectiveness of concerted intervention strategies based on occupant behaviour and personality traits

Abstract

Household electricity consumption influenced by various behavioural intervention strategies is difficult to predict due to the uncertainty that arises from human behaviours and their responses to intervention. Based on an energy conservation experiment conducted in Hangzhou, China, the study aims to develop an improved Support Vector Regression model that is capable of predicting household electricity consumption under multiple intervention strategies. This paper firstly proposes a variable selection approach to determine the best subset of consumption predictors using Akaike Information Criterion. The critical predictors of energy behaviours, personality traits, demographic/building features, weather indicators and the historical monthly consumption is identified in this process. Furthermore, this research also introduces the interaction effect between the energy behaviour and all other predictors to the Support Vector Regression model which applies Gaussian radial basis function optimised by genetic algorithm as the kernel function. More importantly, the proposed model is able to select the optimal intervention strategy and to predict the maximum electricity savings for each household. It enables the households to achieve an average reduction of 12.1% in monthly electricity consumption compared with the conventional behavioural intervention. Moreover, the Monte Carlo simulation is performed to explore the relationship between personality traits, the best-fit intervention strategies and the maximum electricity savings. The study also identifies five types of households with different combinations of extraversion and conscientiousness that respond differently to the optimised interventions. The findings of this study contribute to the residential demand-side energy management by enriching and diversifying personalised behavioural intervention strategies.