Fault diagnosis of energy storage batteries based on dual driving of data and models

Reliable safety warning and fault diagnosis methods for lithium batteries are essential for the safe and stable operation of electrochemical energy storage power stations. Given the current scarcity of failure data for lithium battery storage systems in energy storage power stations and the risks associated with conducting failure experiments on lithium batteries, this paper proposes a method for generating failure data based on the equivalent circuit model (ECM)of lithium-ion batteries. To achieve early fault diagnosis of energy storage batteries, a novel lithium battery fault diagnosis method is introduced, combining a Temporal Convolutional Network and Bidirectional Long Short-Term Memory (TCN-BiLSTM) with the ECM. Firstly, the neural network model is trained using actual normal operation data, and an ECM is constructed. Subsequently, multiple sets of fault data are generated to validate the model's effectiveness. The model's superiority in short-term data prediction and the reliability of fault diagnosis are corroborated through various sets of operational data. Validation results demonstrate that the method proposed in this paper can accurately identify early internal short-circuit faults in lithium batteries.