Volume 52

Day-ahead prediction of household multi-appliance usage trajectory considering individual heterogeneity Wu Jialin, Gao Jiajing, Zhang Yi

Abstract

Demand response is currently an effective strategy to address the mismatch between electricity supply and demand. In residential buildings, there is a wide variety of household appliances with different usage characteristics. It is necessary to be aware of occupants’ daily appliance usage behavior in order to better manage home energy system. In this study, a home energy consumption tracking application is developed for a longitudinal survey based on self-monitoring, during which household appliance usage data of 166 users in Shenzhen are collected. Then day-ahead prediction model for appliance usage trajectories is established. CNN+GRU is used to identify appliance usage characteristics as well as the logical relationships among different appliances. In addition, due to the limited amount of data, this paper utilizes random forest regression for day-ahead prediction for multi-appliance usage states based on occupants’ social attributes, weather parameters and home awake state. It is shown that the predicted multi-appliance usage trajectories are more accurate and logical to user behavior. The results provide a reference for the incentive recommendation mechanisms of different appliances under demand response.

Keywords Household appliances, Panel study, Feature extraction, Trajectory prediction