A novel one dimensional convolutional neural network based data-driven vanadium redox flow battery modelling algorithm

This study proposes an innovative data-driven battery modelling algorithm for vanadium redox flow battery (VRB) in power systems. Unlike the existing battery modelling methods, the proposed algorithm employed the simple but computationally efficient one dimensional convolutional neural network (1D-CNN) technique to learn the nonlinear relationships between VRB current, flow rate, state-of-charge (SOC), and voltage. Compared to the two dimensional CNN, which is widely used in lithium-ion battery modelling and monitoring studies, 1D-CNN eliminates the tedious data re-structuring process and provides better accuracy. Thus, it is more suitable for battery modelling based on one dimensional time series data. Furthermore, 1D-CNN is independent of battery model parameters, allowing it to provide superior modelling performance over the existing electrochemical principle-based and equivalent circuit-based modelling methods that rely on the knowledge of accurate battery model. The validity of the proposed 1D-CNN is verified by experiments.