The invention belongs to the field of power system equipment fault prediction, and particularly relates to a prediction method for frequent device fault signals generation based on Bayesian classification. The method is based on the big data operation of the power grid, and divided in to a sub-optimal device fault prediction method and an optimal fault prediction method according to a data volumelevel; using the sub-optimal device fault prediction method when the data volume is less than N years, wherein N<4; using the optimal fault prediction method when the data volume N is large, wherein N>=4. The sub-optimal device fault prediction method solves a problem of difficulty in prediction caused by insufficiency of data volume, and the optimal fault prediction method integrates a pluralityof power grid operation factors, and predicts fault signals generation condition of the power grid equipment at a high precision.