© 2022 IEEE.This paper presents a comparative investigation of AC, + DC, and - DC corona discharge pulse characteristics using machine learning (ML) algorithms. The corona discharges under different types of excitation are generated via a rod-plane elec-trode system with a constant gap spacing. The corona discharge pulses are recorded using a shunt resistor via an oscilloscope. After noise elimination from the discharge pulses, nine features extracted from the noise-free signals are inputted to several ML models to identify the corona discharges with respect to the voltage types. To increase the performance of a single model, ensemble learning, which is the combination of ML algorithms, is employed. It is observed that the corona discharge types are effectively identified with these features using ensemble learning with a high accuracy rate.