To solve the problems such as reduced power generation efficiency and impeded normal operation caused by faults in photovoltaic arrays,a fault diagnosis model for photovoltaic array is proposed based on GAT.The proposed model leverages discrete wavelet transform and sliding window algorithm to capture post-fault steady-state signals and segment them into sub-intervals,which are treated as graph nodes.Then,the K-nearest neighbor method is used to transform them into graphs.The node-level GAT model can automatically extract fault features from voltage and current graphs using the multi-head attention mechanism.The experimental dataset obtained from the laboratory photovoltaic array is applied for testing the proposed model.The result demonstrates that the GAT model reaches 99.790％in diagnosing various photovoltaic array faults and outperforms other compared network models.
万方数据
维普
