Abstract: Overvoltage recognition is of great importance for the analysis of origins of overvoltages and the induced failures, the improvement of insulation coordination between the power transmission lines and the electrical apparatus.A method to obtain the characteristic quantities of overvoltages is proposed based on the energy distribution of wavelet transform.In view of the eigenvectors overlap, a hierarchical classification method for overvoltage recognition was applied . By constructing a multi-layer overvoltage classification tree, the eigenvector set was selected from different modules and different levels, forming the module-level structure, and constituting the best features of the layer. By introducing the genetic algorithm into fuzzy clustering, the problems were solved about the locality and sensitiveness of the initial condition in fuzzy C-means （FCM） clustering and the convergence rate was improved by combining the global search with the local search and employing the migration strategy. The method was applied to the pattern recognition classification for field overvoltages of the power system, and results show that the method can reduce the error rate of classification greatly, and is effective to recognize overvoltages.