Recognition strategy for pantograph-arc based on ZOA-RBFNN

The pantograph-catenary system, as a vital link in the energy supply chain of electrified railways, has a direct impact on the overall reliability of the railway system. The occurrence of pantograph arc not only affects the quality of power collection by locomotives but also poses significant safety concerns. Therefore, the quest for accurate and efficient methods to identify pantograph arcs is of paramount practical importance. This paper proposes a pantograph arc recognition strategy based on Zebra Optimization Algorithm (ZOA) optimized Radial Basis Function Neural Network (RBFNN). Four groups of current collection experiments were conducted under different operating conditions using a self-developed Pantograph arc experimental simulator; And based on the consideration of zero-rest phenomenon during arc occurrence, the collected current data was selected from the data of the middle half cycle of the power frequency for feature calculation. Principal Component Analysis (PCA) was used to screen the features with higher contribution rate as the recognition basis; Furthermore, the ZOA was used to optimize the RBFNN for training and learning the features of the four groups of experimental, the average test level of the model is 98.5%, and the overall level is above 98%.The testing results of the model verified the generalization feasibility of this strategy; Finally, the superiority of this strategy was demonstrated through comparison with optimization and testing results of other types of algorithms.