Optimizing Residential Energy Resources with an Improved Multi-Objective Genetic Algorithm based on Greedy Mutations

Energy management is increasingly becoming an important issue in face of the penetration of renewable generation and the evolution to smart grids. Home energy management systems are aimed to make the integrated optimization of residential energy resources, taking into account energy prices and end-user's requirements. This paper addresses a residential scenario where energy resources are automatically managed to reduce the overall energy cost while considering a set of user-defined comfort preferences. These energy resources include the grid, shiftable appliances, thermostatic loads, static batteries, electric vehicles, and local energy production. The comfort specifications consist of the time slots where the shiftable appliances are preferred to operate and the temperature ranges desired for the thermostatically controlled loads. The conflicting objectives are addressed by a multi-objective genetic algorithm that aims to minimize the overall energy cost and the user's dissatisfaction. This paper proposes a set of novel operators that result in statistically significant improvements in terms of hypervolume values when compared to a recently proposed genetic algorithm customized to address this same scenario. These novel operators include a different population initialization, a greedy mutation, and two geometric crossovers. The effect of the proposed operators on the resulting allocation of energy resources is analyzed.