Abstract: Partial discharge(PD) is a key diagnostic index to characterize the insulation deterioration state of electrical equipment, which can effectively diagnose insulation defects of different types and gradients of electrical equipment. Aiming at the long-standing problem of insulation diagnosis and evaluation of power equipment, pattern recognition of partial discharge signal and insulation state evaluation are mainly used at present. Traditional classification and recognition methods require experts to manually extract appropriate features from the original data, and then use these features to diagnose the type and severity of PD. At the same time, related artificial intelligence algorithms are gradually developed, and the effective features extracted manually are used as the input of machine learning(ML) algorithm. As a typical artificial intelligence algorithm, deep learning(DL) has better accuracy than traditional ML methods, and provides more effective automatic recognition technology, which is widely used in automatic feature extraction and classification of PD. The research status of DL in partial discharge feature classification and condition assessment is analyzed, and its future development trend is prospected in combination with the development needs of smart grid, smart rail transit and other fields.