Abstract: To address the issues of inaccurate identification and poor suppression effectiveness of ferroresonance overvoltage, this paper proposes a ferroresonance overvoltage classification method based on chaos theory and support vector machines. First, the waveform characteristics of different types of ferroresonance overvoltage are analyzed. Second, by integrating time-domain analysis, frequency-domain analysis, and nonlinear dynamic characteristics, five feature quantities capable of effectively distinguishing sub-harmonic, fundamentalfrequency, high-frequency, quasi-periodic, and chaotic ferroresonance overvoltages are extracted. Finally, the extracted feature quantities are applied to ferroresonance overvoltage identification, and the effectiveness of the proposed method is validated using 35 kV and 10 kV ferroresonance overvoltage waveforms from a distribution network. Simulation and experimental results demonstrate that the three proposed feature quantities improve the identification accuracy of different ferroresonance overvoltage types by an average of 10.2%. The method can accurately identify various ferroresonance overvoltages and exhibits fast identification speed, showing promising application prospects in the diagnosis and mitigation of overvoltage faults in distribution networks.