Photovoltaic power(PV) generation is considered as a renewable energy technology. Due to the nonlinear behav-ior of photovoltaic modules, Maximum power point tracking (MPPT) technology is essential for high-efficiency photovoltaic systems. Although traditional MPPT technologies are simple to implement, they should use trial and error to tune their control parameters, which is the result of unsatisfactory performance in a changing environment. Different from the traditional MPPT technology, the MPPT based on reinforcement learning(RL-MPPT) method has self-learning ability and better applicability in the changing environment. However, the state space size of the method greatly affects the tracking ability. Therefore, this paper compares the tracking performance of the three state spaces through the results of simulation and experiment under En50530 test procedure.