Modeling of All-Vanadium Redox Battery based on CNN-BiGRU Neural Network

The operation process of all-vanadium redox batteries (VRBs) involves coupling of multiple physical fields such as flow field, electrochemical field, and thermal field. Traditional modeling methods are complex, time-consuming, and with low accuracy. An innovative modeling method based on CNN-BiGRU neural network is proposed in this paper, which selects flow rate, current, and state of charge (SOC) as input features and directly outputs the terminal voltage of the battery. Experimental results validate the accuracy of the model, with a mean absolute error (MAE) as low as 13.2 mV and a mean absolute percentage error (MAPE) is as low as 0.96% under multiple operating conditions with flow rate variations. The model demonstrates extraordinary accuracy and wide applicability.