A novel U-Net based data-driven vanadium redox flow battery modelling approach. (10th March 2023)

• Record Type: Journal Article
• Authors: Li, Ran; Xiong, Binyu; Zhang, Shaofeng; Zhang, Xinan; Li, Yifeng; Iu, Herbert; Fernando, Tyrone
• Is Part Of: Electrochimica acta. Volume 444(2023)
• Publication Date: 2023-03-10
• Subjects: Vanadium redox flow battery -- Modelling -- U-Net -- Data-driven; Electrochemistry -- Periodicals; Electrochemistry, Industrial -- Periodicals; 541.37
• DOI: 10.1016/j.electacta.2023.141998
• Languages: English
• ISSNs: 0013-4686

Abstract

Highlights: The proposed approach employs a data-trained U-Net to accurately describe the nonlinear dynamics of VRB electrical variables, and the technique does not require any empirical parameter adjustment. Unlike most of the existing methods, the influence of varying input flow rate is considered, which fundamentally improves model accuracy. It is theoretically model parameter independent and can be utilized directly by electrical engineers with no electrochemical background. U-Net is well-known for its high computational efficiency in training. More importantly, the trained U-Net is mathematically simple, allowing it to be easily incorporated into numerical simulation models of power system. Abstract: This research proposes a highly accurate data-driven vanadium redox flow battery (VRB) modelling approach for power engineering applications. The proposed approach addresses the common problem of excessive model dependency in the existing electrochemical principle or equivalent circuit based VRB modelling methods. Furthermore, an experimentally trained U-Net is applied to directly learn the behavioral relationship between VRB current, flow rate, state-of-charge, and voltage with excellent accuracy, avoiding the usage of model parameters that are subject to variations. Once trained, the U-Net based neural network becomes mathematically very simple and thus, can be easily implemented in simulation studies. This contributes to substantially simplify the analysis of electrical systems with VRB. The validity of the proposed approach is verified experimentally.