Minor Fault Diagnosis of Lithium Battery with DOA-VMD-RNTransformer Hybrid Model

The rapid development of new energy sources has led to an increasing reliance on lithium-ion batteries (LIBs). However, the complex operational conditions of LIBs make batteries susceptible to minor failures, jeopardizing the safety and stability of batteries. This paper investigates multiple minor faults of LIBs. Therefore, a DOA-VMD-RNTransformer hybrid model is proposed, First, dream optimization algorithm (DOA) optimizes the parameters of variational mode decomposition (VMD), which is convenient for the fault feature extraction of battery time series data. Subsequently, a hybrid model based on recurrent neural network (RNN) and Transformer (RNTransformer) is established to minor fault diagnosis. The experimental results demonstrate that the proposed method can achieve a diagnosis accuracy of 98.45%, which realizes high accuracy and robustness in diagnosing various minor faults of batteries. Finally, the ablation study and the comparison with the existing methods confirm the superiority and feasibility of the proposed method.