Abstract: To address the shortcomings of the traditional extreme learning machine in PV fault diagnosis, a method using the Chaotic and golden sine bald eagle search (CGSBES) algorithm to optimize the input weights and hidden layer neurons of the extreme learning machine (ELM) is proposed, and the CGSBES-ELM model is proposed to diagnose faults in PV modules. By analyzing the I-V curve and P-V curve changes in the fault state of a 10kW photovoltaic module simulation model, the fault feature quantities were extracted and a fault diagnosis model was established. The proposed fault diagnosis model was validated based on actual photovoltaic data. The experimental results show that the CGSBES-ELM model can accurately identify the fault types of PV modules using the 6-dimensional fault feature vector and has a higher fault diagnosis accuracy.