Fault Diagnosis of Large Grid With Alexnet Neural Network Based on Multi-source Data Fusion

In allusion to the diagnosis of power grid faults occurred in the extreme conditions such as typhoons, freezing and so on, a scheme, in which firstly an Alexnet model was trained by the data from standard telecommand and wide area measurement system (abbr. WAMS) and then the trained Alexnet was applied to power grid fault diagnosis, was proposed. Firstly, the standard fault remote signaling information and WAMS data were utilized to construct the input image matrix to train the Alexnet. Secondly, the high dimensional feature extraction method of Alexnet input image was analyzed and a method to construct the input image matrix with optimal distribution structure was put forward and a fault diagnosis model was formed. Finally, taking Hainan Island power grid subjected to typhoon attack as the scene a simulation model was built to verify the Alexnet fault diagnosis model.