Abstract: A reinforcement learning gaining-sharing knowledge-based algorithm（RLGSK） is proposed to accurately identify the parameters of solar cell models. The selection mechanism of the junior and senior phases of gaining-sharing knowledge-based algorithm（GSK） is too rigid, which makes GSK difficult to fully balance the global and local searches when dealing with the solar cell model parameter identification. Besides, it has some problems such as slow convergence and low accuracy. In this regard, on the one hand, the ratio of individuals in the junior and senior phases during iteration is adjusted by reinforcement learning to achieve flexible adjustment of knowledge gaining and sharing in different scenarios. On the other hand, a performance-oriented population size reduction technique is used to achieve efficient use of computational resources and improve algorithm performance. RLGSK is applied to five cases and compared with other algorithms. The results show that the search accuracy, stability and convergence speed of RLGSK are greatly improved compared to GSK, and it is also very competitive with other algorithms.