Residential lighting load profile modelling: ANFIS approach using weighted and non-weighted data

Abstract

This study involves the use of adaptive neural fuzzy inference system (ANFIS) for residential lighting load profile development and evaluation of energy and demand side management (DSM) initiatives. Three variable factors are considered in this study namely, natural light, occupancy (active), and income level. A better correlation of fit and reduced root mean square error was obtained after validation of the developed model using the investigative data—weighted and non-weighted approach (natural lighting). The technique showed that income level of the class in relation to the area (location), working lifestyle of individuals in relation to behavioural pattern, and effect of natural lighting are highly essential and need to be incorporated in any load profile development. The generalisation of income needs to be revisited; emerging middle and realised middle-income predictors have shown that their behavioural pattern differs. Forecast based on averages of lamps per households from a survey of an income class to determine lighting usage is prone to high errors. The developed methodology of the ANFIS gives better lighting prediction accuracy in accordance with the learning characteristics of light usage complexities.