Multi-step depth model predictive control for photovoltaic maximum power point tracking under partial shading conditions

Abstract

This work avoids the photovoltaic array operating point falling into local maximum power points (MPPs) and optimizes the dynamic process of photovoltaic array global MPP tracking under partial shading conditions (PSCs) by the proposed multi-step depth model predictive control (MDMPC). The MDMPC updates the dynamic conductance at different moments to achieve multi-step accurate predictions of the system state. Combining deep neural networks and the global MPP calculation model can rapidly estimate the global MPP under different topologies. The proposed two-stage optimization adaptively adjusts the optimization function according to the purpose of different stages. The simulated system response curves and performance indexes under different PSCs show that the MDMPC can avoid the operating point falling into local MPPs and accurately rapidly track to the global MPP with higher power generation efficiency. In addition, the hardware-in-the-loop experimental results verify that MDMPC can rapidly and accurately track the global MPP under three PSCs.