基于人工智能技术的钒氧化还原液流电池

Abstract Vanadium redox flow battery (VRFB) offers a sustainable and reliable solution for large-scale energy storage applications. This study represents the first investigation into the comprehensive data-driven analysis of inter-parameter correlation and prediction of the energy efficiency of VRFBs utilizing the Gaussian Process Regression (GPR) model. Namely, 420 VRFB datasets were collected from the literature, whereas 10 structural and 2 operational features are considered input parameters. Indeed, in the VRFB cells with the greater active area, i.e., pilot-to-commercial-scale applications, the Serpentine flow field configuration, higher electrolyte concentration, thicker electrodes, and higher felt compression are more prevalent. The outcomes reveal that the current density, membrane type, and electrode treatment with the respective Pearson correlation coefficient values of −0.4167, 0.2862, and 0.1546 significantly affect the VRFBs' energy efficiency. Besides, the developed ML models can accurately result in the associated energy efficiency in the VRFBs, with the highest accuracy of the GPR- Matern5/2. The training and testing R 2 values are 0.9933 and 0.9565, respectively, indicating near-perfect accuracy, making it a reliable model. This research paves the way for improving VRFB performance, advancing its practical application, and providing key insights into AI-driven battery design.