Neural Network Model with Cascaded Structure for State-of-charge Estimation in Lithium-ion Batteries

In this article, a new methodology is proposed to achieve high-accuracy estimation of battery state-of-charge (SOC). In this method, the AdaBoost-AOA-BPNN model employs an arithmetic optimization algorithm (AOA) to optimize the threshold values and initial weights of the backpropagation neural network (BPNN). In addition, this model is employed as a strong learner, leveraging the AdaBoost technique to integrate multiple AOA-BPNN sub-models as weak learners. The final prediction output is obtained by the sum of the output of each sub-model/weak learner multiplied by a corresponding weighting factor. The weighting factor is adaptively adjusted by assigning greater importance to highly correlated samples and reducing the significance of samples with low correlation. This paper presents theoretical discussions and findings from a proof-of-concept experimental model to assess and verify the effectiveness of the AdaBoost-AOA-BPNN framework.