Decentralized stochastic control for building energy and comfort management

The problem of building energy and comfort management involves complex decision making in the presence of uncertainties. Therefore, it is required to model the problem in terms of a decision process that can handle uncertainties. Such model is developed in this paper using Markov decision process. The proposed model incorporates simultaneous comfort and energy management with incorporation of occupant dynamics and uncertainties. Due to high complexity of the problem, it is not practical to directly solve the resulting Markov decision process. Therefore, a distributed approach has been proposed that decomposes the problem in two dimensions. First, the problem is decomposed with respect to thermally and visually isolated zones in the building i.e. halls, rooms, closed offices etc. Second, within each isolated zone, the problem is decomposed with respect to comfort goals, i.e., thermal comfort, visual comfort, humidity and air quality. It has been shown that the reduction in computational complexity is achieved with minimal loss in optimality. This is possible due to naturally existing conditional independence among the random variables involved in problem. A comparison of the proposed approach with recent approaches has been presented that indicates that the proposed approach offers more features than any of the existing approaches. A demonstrative case study has also been included to show how the proposed approach can be implemented and how it is expected to behave. The results are encouraging towards practicality of the proposed approach.