Mean-field control for energy efficient buildings

In this paper, we consider the problem of distributed set-point temperature regulation in a large building. With a large number of zones, the problem becomes intractable with standard control approaches due to the large state space dimension of the dynamic model. To mitigate complexity, we develop here a mean-field control approach applicable to large-scale control problems in buildings. The mean-field here represents the net effect of the entire building envelope on any individual zone. Rather than solving the large-scale centralized problem, we explore distributed game-theoretic solution approaches that work by optimizing with respect to the mean-field. The methodology is illustrated with a numerical example in a simulation environment.