Abstract: Photovoltaic power generation as an important component of a new type of power electronic system,has random fluctuations.And how to improve its power supply efficiency and reliability becomes a challenging problem.For this reason, this paper proposes a dual-mode maximum power point tracking (MPPT) algorithm based on partial shading detection. The photovoltaic power prediction model based on long short-term memory neural network (LSTM) is used to detect partial shading conditions. On this basis, the proposed dual-mode MPPT algorithm considers both normal system operation and localized shading conditions. Based on radial basis function neural network (RBFNN), MPPT algorithm under partial shading is implemented, while the traditional perturbation observation method is used to implement MPPT algorithm under normal operation conditions. In order to validate the effectiveness of the proposed MPPT method, a PV system model is built and the performance of the proposed algorithm is analyzed for different operating conditions in the MATLAB/Simulink environment.

Key words: maximum power point tracking (MPPT), radial basis function neural network (RBFNN), power prediction, long short-term memory (LSTM), photovoltaic power generation