Addressing the stochastic nature of energy management in smart homes

In the future, automated smart home energy management systems (SHEMSs) will assist residential energy users to schedule and coordinate their energy use. In order to undertake efficient and robust scheduling of distributed energy resources, such a SHEMS needs to consider the stochastic nature of the household's energy use and the intermittent nature of its distributed generation. Currently, stochastic mixed-integer linear programming (MILP), particle swarm optimization and dynamic programming approaches have been proposed for incorporating these stochastic variables. However, these approaches result in a SHEMS with very costly computational requirements or lower quality solutions. Given this context, this paper discusses the drawbacks associated with these existing methods by comparing a SHEMS using stochastic MILP with heuristic scenario reduction techniques to one using a dynamic programming approach. Then, drawing on analysis of the two methods above, this paper discusses ways of reducing the computational burden of the stochastic optimization framework by using approximate dynamic programming to implement a SHEMS.