Abstract: At present, the high-frequency transient overvoltage caused by the frequent operations on electrical equipment or faults in offshore wind farms, is particularly severe. In order to analyze the transient characteristics of internal overvoltage in offshore wind farms, this paper firstly proposed a signal feature extraction method based on mathematical morphology, constructing a new morphological structure operator and utilizing multi-scale mathematical morphological decomposition to extract the high and low frequency information of transient overvoltage. Then two time-domain identification feature indexes were constructed for identifying the type of transient overvoltage in offshore wind farms. Finally, based on the high frequency feature index and the high and low frequency energy rate feature index proposed by this paper, different types of internal transient overvoltage classification could be classified using the support vector machine classifier model. The results show that compared with the traditional Wavelet algorithm, the feature extracted by the proposed mathematical morphology algorithm had a clearer degree of discrimination, which could accurately identify the overvoltage types, laying a foundation for the protection setting and insulation coordination for the electrical equipment in the offshore wind farm substations.