Stochastic optimal dispatch of PV/wind/diesel/battery microgrids using state-space approximate dynamic programming

In the operation of microgrids (MGs), the stochastic production of solar/wind, the discrete variables of photovoltaic (PV) inverter's auxiliary service state and diesel generators' (DGs') off-on state generally need to be considered, and a stochastic mixed-integer non-linear non-convex programming (MINNP) model is established for the optimal dispatch of MGs. In this model, the expected value of the sum of DGs' operation as well as start-up cost, the network-loss cost and the PV inverter's auxiliary service cost, is considered as the objective function. The stochastic MINNP model is transformed into a stochastic mixed-integer second-order cone programming (MISOCP) model to reduce the computational complexity. The state-space approximate dynamic programming algorithm is adopted to solve the stochastic MISOCP model. In the algorithm, based on the approximate value functions of typical states that are computed according to the Markov decision process, solving the optimisation model of multiple periods is executed by solving each period's optimisation model one by one to improve the computational efficiency. Meanwhile, parallel computing is executed to greatly improve the efficiency of the proposed algorithm. Test results on two modified IEEE-33 bus and IEEE-123 bus islanded MGs with PV/wind/diesel/battery demonstrate the correctness and efficiency of the proposed model and algorithm.