The output power of photovoltaic (PV) arrays exhibits strong randomness and volatility. In the event of a fault, it can severely impact the safety and stable operation of the power system. Addressing the challenges of low accuracy and slow convergence in current PV fault diagnosis, this paper proposes a PV array fault diagnosis method based on the grasshopper optimization algorithm-support vector machine (GOA-SVM) model. Firstly, an equivalent circuit model of the PV array is established to analyze the variation characteristics of the PV array's voltage-current curve. Secondly, considering environmental factors and the nonlinear changes in the scale of the PV array, feature quantities reflecting different fault characteristics are extracted, and the data is mapped into a high-dimensional space for nonlinear processing. Finally, an improved method for optimizing the nonlinear support vector machine using GOA is proposed, and a GOA-SVM PV array fault diagnosis model is established, with simulations conducted using practical examples. The research results indicate that this method can be applied to various PV array models of different scales and effectively diagnose faults in PV arrays. For a 4×3 PV array scale, the data simulation classification accuracy can reach 99.8088%. When validated using the publicly available dataset from the national institute of standards and technology (NIST), the fault diagnosis accuracy achieves 92.3682%. Compared with other methods, this approach demonstrates significant improvements in recall rate and F1-Score.