The super-twisting sliding mode control (ST-SMC) is widely used in fuel cell system control due to the simple control law and strong robustness. However, it requires an accurate system model. Generally, literature usually reduces the system order and directly gives the empirical value for the system and controller parameters or conducts coefficient identification of the fuel cell voltage model, lacking the specific identification method for system physical parameters and controller parameters. In this paper, a relatively complete control-oriented nine-state fuel cell system model was established, including the model of compressor flow using the artificial neural network method, and the improved voltage model. Then, the data-driven method for key parameter identification was proposed, including the fuel cell throttle factor and motor voltage changing rate considering time delay. In addition, the parameter tuning method for controller design was proposed as well. These two methods are of originality. After the model validation in the perspective of steady and transient performance, the comparison was carried out between the ST-SMC and PID controller. It is found that the throttle factor of the cathodic fuel cell inlet and the delay effect in terms of changing rate of motor voltage impact the system model, where the throttle factor is time-variant, and the delay is noticeable which differs with the step magnitude of the motor voltage. The parameter tuning and boundary estimation of ST-SMC are very specific, owing to the process of treating flow rate as a state, not speed, and are convenient to be generalized. The better ability in anti-flooding exhibits the importance of parameter identification. Although the study is conducted in a low-pressure system, the method proposed in this paper is universal and could be applied to other fuel cell controls for better system efficiency and reliability.

超扭滑模控制（ST-SMC）由于控制规律简单，鲁棒性强，被广泛应用于燃料电池系统控制。但是，它需要一个准确的系统模型。通常，文献通常将系统阶数降低，直接给出系统参数和控制器参数的经验值，或者对燃料电池电压模型进行系数辨识，缺乏系统物理参数和控制器参数的具体辨识方法。本文建立了一个比较完整的面向控制的九态燃料电池系统模型，包括采用人工神经网络方法的压缩机流量模型和改进的电压模型。然后，提出了数据驱动的关键参数识别方法，包括燃料电池节流系数和考虑时间延迟的电机电压变化率。此外，还提出了控制器设计的参数整定方法。这两种方法都是独创的。在从稳态和瞬态性能的角度进行模型验证后，对 ST-SMC 和 PID 控制器进行了比较。发现阴极燃料电池入口的节流因子和电机电压变化率的延迟效应对系统模型有影响，其中节流因子是时变的，并且延迟是明显的，随着步长的不同而不同电机电压。ST-SMC 的参数调整和边界估计是非常具体的，由于将流量视为状态而不是速度的过程，并且便于推广。较好的抗洪能力体现了参数辨识的重要性。尽管该研究是在低压系统中进行的，但本文提出的方法是通用的，可以应用于其他燃料电池控制，以提高系统效率和可靠性。