Improved approximate dynamic programming for real-time economic dispatch of integrated microgrids

Economic dispatch of electricity-heat microgrid is critical for real-time power generation and storage. However, conventional economic dispatch algorithms are generally integrated with static unit models without considering dynamics of units, thus leading to difficulties for real deployment in stochastical environments. In this paper, we propose a novel approximate dynamic programming (ADP) based real-time optimization algorithm. Specifically, the proposed ADP is employed to solve the Markov decision process with considering the dynamic process of combined-cycle gas turbine. Furthermore, we also design a novel weighted piecewise linear function to achieve the near-optimal solution, which is simple but effective for computational complexity reduction. In the experimental section, we conduct extensive experiments with comparisons to other economic dispatch methods. The experimental results indicate that: 1) The dynamic process of energy conversion brings more practical solutions; 2) The proposed ADP-based method could handle the stochasticity of the microgrid; 3) The proposed method outperforms the other intra-day optimization policies in both economical and computational efficiency.