A comparison between two MPC algorithms for demand charge reduction in a real-world microgrid system

This paper describes an evaluation between two model predictive control (MPC) algorithms for microgrid energy management combined with solar production and battery energy storage for demand charge reduction in a real-world microgrid system. The first control algorithm is a constant threshold MPC (CT-MPC) that works well on a system with relatively stable solar generation and a well-known building load profile. CT-MPC can maintain the on-peak demand under a certain value during the entire on-peak rate period. The second control algorithm is an adjusting demand threshold MPC (ADT-MPC). ADT-MPC can better deal with unpredictable solar generation and/or changing building loads. The on-peak threshold under this algorithm is adjusted to the optimal value during the on-peak rate period. As expected, The CT-MPC algorithm performs well when coupled with accurate forecast models while the ADT-MPC algorithm excels when forecasting is more unpredictable.