Fuzzy predictive filtering in nonlinear economic model predictive control for demand response

The performance of a model predictive controller (MPC) is highly correlated with the model's accuracy. This paper introduces an economic model predictive control (EMPC) scheme based on a nonlinear model, which uses a branch-and-bound tree search for solving the inherent non-convex optimization problem. Moreover, to reduce the computation time and improve the controller's performance, a fuzzy predictive filter is introduced. With the purpose of testing the developed EMPC, a simulation controlling the temperature levels of an intelligent office building (PowerFlexHouse), with and without fuzzy filtering, is performed. The results show that the controller achieves a good performance while keeping the temperature inside the predefined comfort limits. Fuzzy predictive filtering has shown to be an effective tool which is capable of reducing the computational burden and increasing the performance level of the control algorithm.