Uncertainty aware minority game based energy management system for smart buildings

For the sake of accurate energy resource allocation in smart buildings with hybrid solar energy and main electrical grid, an uncertainty-aware minority-game based energy management system (UAMG-EMS) is introduced in this paper. Multiple agents are deployed in the building, and are able to consider two types of uncertainties: (i) stochastic noise from energy meters/sensors; and (ii) uncertain working behaviors from load side. Firstly, agents can perform Kalman Filter based error-correction algorithm to reduce the stochastic noise coming from energy meters/sensors. Moreover, agents can have supervised learning to predict uncertain energy profiles. Afterwards, agents can play a modified minority game based energy management to allocate the limited solar energy resource. To extend the scalability of agents for the entire building, K-means based classifier is applied to characterize the types of agents and hence can reduce the number of agents for large-scale buildings. Compared with the conventional minority-game based energy management system (MG-EMS) without considering uncertainty, our UAMG-EMS shows about 37% reduction of unbalance in fair solar energy allocation, and also about 23% reduction of noise influence merely based on inaccurate energy meters/sensors.