考虑质子交换膜燃料电池(Proton exchange membrane fuel cell,PEMFC)的净输出功率最优化跟踪问题,提出一种基于显式模型预测控制(Explicit model predictive control,EMPC)的无偏优化控制策略,以实现PEMFC系统在面对模型失配和约束条件的影响下,对期望净输出功率的鲁棒稳态无偏差跟踪,以及将系统工作点调节在最(次)优工作曲线附近,达到效率最优化.首先,对PEMFC系统的非线性特性进行线性化,推导出面向控制器设计的线性控制模型.在此基础上,通过设计状态观测器,实现对状态变量和模型失配度的估计;通过设计目标计算器,实现对最(次)优净输出功率的跟踪和空气压缩机工作边界的约束;通过设计EMPC控制器,对空气压缩机工作边界约束进行松弛,以获得可行的显式解,从而构成无偏EMPC控制策略.最后,搭建硬件在环测试(Hardware in loop,HiL)平台,基于标准的车用燃料电池堆试验循环工况,通过与传统MPC和EMPC的仿真试验对比,验证所提出的无偏EMPC控制策略可将系统整体工作效率从41％提高到47.6％,输出功率的平均跟踪误差绝对值从0.26 kW降低到0.17 kW,单步平均计算耗时从22.2 ms缩短到2.1 ms.
Offset-free Explicit Model Predictive Control and Its Application on the Net Output Power Tracking of PEMFC System
Considering the optimal tracking issue of the net output power for the proton exchange membrane fuel cell(PEMFC),an explicit model predictive control(EMPC)based offset-free optimal control strategy is proposed,which aims at achieving the robust steady state offset-free tracking of desired net output power,in terms of system's operating constraints and model mismatching,and is for the purpose of optimizing the system efficiency,by regulating the PEMFC's operating points at around the optimal(sub-optimal)operating curve.Firstly,a control-oriented linear model is proposed by linearizing the PEMFC system.Then,a state observer is designed to estimate the state variables and model mismatching.A target calculator is designed to track the optimal net output power and to constrain the operating boundary of the air compressor.An EMPC controller is designed to relax the air compressor boundary constraints and achieve a reachable explicit solution,and consequently an offset-free EMPC strategy is proposed involves the above three parts.Finally,a hardware in loop test bench is set up,and a simulation is performed based on a standard vehicular fuel cell stack test cycle,in comparing with the conventional MPC and EMPC strategy.The results demonstrate that the off-set free EMPC control strategy is able to improve the system efficiency from 41％up to 47.6％,and to reduce the absolute average error of the output power tracking from 0.26 kW down to 0.17 kW,and the single step average time cost is reduced from 22.2 ms down to 2.1 ms,respectively.
proton exchange membrane fuel cellnet output power trackingrobust optimizationoffset-free explicit model predictive control
NETL
NSTL
万方数据
