Abstract: This study combines the Bidirectional Long Short-Term Memory Network (Bi-LSTM) with statistical methods based on the Weibull distribution to improve the accuracy of condition assessment of oil-immersed transformers and predict their remaining useful life. It systematically summarizes data-driven methods for equipment fault condition assessment and life prediction, determines characteristic state transition sequences of transformers, constructs a Bi-LSTM model for fault assessment, establishes a Weibull-based life prediction model, and validates the effectiveness with actual data.