Control of a networked microgrid system with an approximate dynamic programming approach

Constructing microgrids with renewable energy systems and storage units can be a feasible solution to smooth the fluctuation of renewable power and increase its penetration. In recent years, research on microgrids has proven that the overall electricity and operational costs in microgrids can be further reduced with a networked microgrid system. One major issue in controlling microgrids with renewable power systems is the uncertainties due to forecasting errors, and optimization problems in a microgrid network can be could be challenging to implement due to a large number of uncertainties. In this paper, we propose an approximate dynamic programming-based power management strategy in a microgrid network system, and the objective is to minimize the overall cost over a control horizon with predicted renewable power, electricity price and load demand. The network consists of several microgrids and a centralized energy management system. The use of the approximate dynamic programming algorithm is to address the 'curse of dimensionality' due to high dimensional state and control spaces in the network.