Fault Diagnosis of PV Inverter Open-Switch Fault with Low Sampling Frequency Data

As a critical component in photovoltaic (PV) systems, inverters are prone to failure, yet their fault diagnosis is often constrained by low sampling frequency, resulting in sparse data and significant information loss. To address these challenges, this paper proposes a deep residual network integrating multi-scale temporal modeling and adaptive attention mechanisms. Experimental results indicate that the proposed method achieves consistently high diagnostic accuracy across all sampling intervals, with improvements of 9.1% under the lowest sampling frequency condition. Thereby providing a robust and practical solution for low-cost fault detection in PV system deployment.