基于模型自学习的燃料电池空气系统自抗扰控制

Active disturbance rejection control of fuel cell air system based on model self-learning

中文关键词  质子交换膜燃料电池  模型训练  阶次选择  自抗扰控制

英文关键词  proton exchange membrane fuel cells (PEMFC)  model training  order selection  active disturbance rejec tion control

中文摘要

      面向控制的燃料电池空气系统模型存在工况适应范围小的问题,因此,系统时变后导致的模型不准确将恶 化基于模型的控制效果.对此,本研究开发了一种基于贝叶斯学习模型的自抗扰控制策略,以实现宽工况范围和长 寿命周期内的精确控制.首先,建立面向控制的燃料电池空气系统四阶模型,采用贝叶斯估计,根据不同工况下燃料 电池系统的测试数据对模型参数进行训练,并借助极大似然估计选择空压机流量多项式模型的适宜阶次,降低模型 参数时变造成的影响,增大模型的工况适应范围.此外,针对空气系统流量与压力的耦合问题,设计了自抗扰解耦控 制方案,将耦合效应视为系统总扰动,利用扩张状态观测器实时估计并补偿至控制律中,实现精准解耦.最后,基于 MATLAB/Simulink环境搭建了燃料电池空气系统仿真平台,仿真结果显示,该方法有效确保了模型高精度,实现了 流量与压力的精确协同控制,保障了燃料电池系统的安全高效运行.

英文摘要

      The control-oriented model of the fuel cell air system exhibits limited adaptability to operating conditions, thereby exacerbating the inaccuracies caused by temporal changes in the system and subsequently compromising its control effectiveness. To overcome this, a study presents an active disturbance rejection control approach rooted in Bayesian learn ing, aimed at precise control across broad operating conditions and longevity. A fourth-order model tailored for fuel cell air systems is formulated, with Bayesian estimation refining parameters using test data from diverse working conditions. Maximum likelihood estimation selects the optimal polynomial order for the air compressor flow model, mitigating the impact of time-varying parameters and enhancing model adaptability. Addressing flow-pressure coupling, an active dis turbance rejection decoupling control is introduced. Treating coupling as systemic disturbance, an extended state observer estimates and compensates in real-time, enabling precise decoupling. A MATLAB/Simulink simulation platform validates the method. Results highlight its effectiveness in ensuring model accuracy, facilitating precise flow-pressure control, and ensuring safe, efficient fuel cell system operation.