The framework of a multi‐scale model that couples a deep neural network, a widely used machine learning approach, with a partial differential equation solver and provides understanding of the relationship between the pore‐scale electrode structure reaction and device‐scale electrochemical reaction uniformity within a redox flow battery is introduced. A deep neural network is trained and validated using 128 pore‐scale simulations that provide a quantitative relationship between battery operating conditions and uniformity of the surface reaction for the pore‐scale sample. Using the framework, information about surface reaction uniformity at the pore level to combined uniformity at the device level is upscaled. The information obtained using the framework and deep neural network against the experimental measurements is also validated. Based on the multi‐scale model results, a time‐varying optimization of electrolyte inlet velocity is established, which leads to a significant reduction in pump power consumption for targeted surface reaction uniformity but little reduction in electric power output for discharging. The multi‐scale model coupled with the deep neural network approach establishes the critical link between the micro‐structure of a flow‐battery component and its performance at the device scale, thereby providing rationale for further operational or material optimization.

中文翻译：

---

氧化还原液流电池的机器学习耦合多尺度建模

多尺度模型的框架，该模型将深层神经网络，一种广泛使用的机器学习方法与偏微分方程求解器结合在一起，可以理解孔尺度电极结构反应与装置尺度电化学反应均匀性之间的关系引入了氧化还原液流电池。使用128个孔尺度模拟对深度神经网络进行了训练和验证，这些模拟提供了电池工作条件与孔尺度样品表面反应的均匀性之间的定量关系。使用该框架，有关孔隙水平表面反应均匀性到器件水平组合均匀性的信息得到了放大。使用框架和深度神经网络针对实验测量获得的信息也得到了验证。基于多尺度模型结果，建立了电解质入口速度随时间变化的优化方法，从而显着降低了目标表面反应均匀性所需的泵功率消耗，但减少了放电功率输出。多尺度模型与深度神经网络方法相结合，在流动电池组件的微观结构与其在设备规模上的性能之间建立了关键的联系，从而为进一步的操作或材料优化提供了依据。