Vanadium flow batteries(VFBs)are well suitable for grid-scale energy storage owing to their long lifespan,high efficiency and safety.State of charge(SOC)monitoring is essential for battery health assessment and system...Vanadium flow batteries(VFBs)are well suitable for grid-scale energy storage owing to their long lifespan,high efficiency and safety.State of charge(SOC)monitoring is essential for battery health assessment and system management.However,accurate SOC determination during operation remains challenging due to vanadium ion crossover and side reactions that disrupt the valence and concentration balance between positive and negative electrolytes.Herein,an inverted transformer model,namely iTransformer,is employed to predict the SOC in VFB systems during charge–discharge cycles.The iTransformer-SOC model can achieve high accuracy and robustness.Even with training limited to the first three cycles,the model predicts SOC for the next 21 cycles with mean absolute percentage error(MAPE)less than 0.03.It can adapt to power variations and electrolyte rebalancing scenarios.Most importantly,an iTransformer-based SOC monitoring system was validated and confirmed by a 10 kW VFB system,demonstrating superior performance in predicting SOC of complete charge–discharge cycles(MAPE<0.05,less than 1/3 of the traditional open-circuit voltage(OCV)method's error).This datadriven approach provides a robust framework for real-time SOC monitoring in VFB systems,serving as a complementary alternative to physics-based model without requiring prior knowledge of system dynamics.展开更多
基金supported by the Key R&D Projects of the National Natural Science Foundation of China(2022YFB2404904)the National Natural Science Foundation of China(22309178)+1 种基金the Strategic Priority Research Program of the CAS(XDA0400402)the Liaoning International Cooperation Project(2023JH2/10700002)。
摘要Vanadium flow batteries(VFBs)are well suitable for grid-scale energy storage owing to their long lifespan,high efficiency and safety.State of charge(SOC)monitoring is essential for battery health assessment and system management.However,accurate SOC determination during operation remains challenging due to vanadium ion crossover and side reactions that disrupt the valence and concentration balance between positive and negative electrolytes.Herein,an inverted transformer model,namely iTransformer,is employed to predict the SOC in VFB systems during charge–discharge cycles.The iTransformer-SOC model can achieve high accuracy and robustness.Even with training limited to the first three cycles,the model predicts SOC for the next 21 cycles with mean absolute percentage error(MAPE)less than 0.03.It can adapt to power variations and electrolyte rebalancing scenarios.Most importantly,an iTransformer-based SOC monitoring system was validated and confirmed by a 10 kW VFB system,demonstrating superior performance in predicting SOC of complete charge–discharge cycles(MAPE<0.05,less than 1/3 of the traditional open-circuit voltage(OCV)method's error).This datadriven approach provides a robust framework for real-time SOC monitoring in VFB systems,serving as a complementary alternative to physics-based model without requiring prior knowledge of system dynamics.