Learning-Based Energy Management System for Scheduling of Appliances inside Smart Homes

Improper designs of the demand response programs can lead to numerous problems such as customer dissatisfaction and lower participation in these programs. In this paper, a home energy management system is designed which schedules appliances of smart homes based on the user’s specific behavior to address these issues. Two types of demand response programs are proposed for each house which are shifting-based and learning-based programs for shiftable and heating, ventilation and cooling appliances, respectively. The current structure uses machine learning techniques to design the best demand response programs for heating, ventilation and cooling devices of each user based on his/her behavior and desired comfort level. Doing so, the home energy management system is able to achieve energy cost and consumption reduction without causing dissatisfaction and discomfort to the users. Results demonstrate that by using this structure, energy cost and consumption are reduced by 20.32% and 27%, respectively for a single house located in the Austin, Texas area, in one day. The proposed home energy management structure is tested on three additional houses to show the effectiveness of it. Moreover, comparisons with other methods are performed to clarify the benefits of this structure over other methods. The proposed structure is formulated as a mixed-integer linear model with its optimization performed in the General Algebraic Modeling System environment. CPLEX solver is used to solve the optimization problem.