近年来，车载质子交换膜燃料电池（Proton Exchange Membrane Fuel Cell, PEMFC）电堆功率密度、低温冷启动、耐久性等性能指标不断攀升，燃料电池发动机（Fuel Cell Engine, FCE）输出功率不断增大（最高可达100~120 kW），燃料电池汽车动力系统构型逐步由增程式向全功率式发展。全功率燃料电池汽车所搭载的燃料电池发动机功率较大、动力电池容量较小。日本丰田2015年发布的Mirai是典型代表。在这种动力构型中，燃料电池发动机需要满足车辆的大部分功率需求。在燃料电池氢-空-水热各个子系统中，由于空气系统功率消耗大、响应时间慢，且氧气在电堆阴极侧扩散阻力大，容易在燃料电池发动机大范围的功率变化动态过程中出现局部缺气。因此，需要设计合理的动力系统协调控制策略，确保车辆的动力性、经济性和耐久性。本文聚焦全功率燃料电池汽车预测性分层协调控制策略。该策略由两个部分组成，上层预测性自适应缓慢变载运行（Predictive Adaptive Softrun, PAS）能量管理策略、下层空气系统解耦控制策略。其基本理念为：通过LSTM算法提前预知未来某时间段的整车平均功率需求，基于该平均功率协调控制空气系统空压机和背压阀，使燃料电池发动机既能保持一定程度的稳定输出，又能随着整车功率需求动态变化自适应调整其功率输出点。论文首先建立了燃料电池汽车动力系统分层模型，尤其考虑了FCE空气系统动态特性。其次，设计了空气系统的双输入双输出控制算法并验证其有效性。第三，设计了基于LSTM算法的预测性自适应缓慢变载运行PAS策略，针对策略中的三个关键参数进行了寻优。第四，在NEDC工况和实际工况（包括北京-延庆和北京-张家口两段）下，基于相同的空气系统控制策略对比PAS算法和常见的PF（Powerfollow）算法和SR（Softrun）算法。研究结果表明：在NEDC工况下，由于整车平均需求小且工况可重复性好，存在一种SR算法在百公里氢耗方面优于PAS算法；在实际工况下，由于整车平均需求功率大且工况随机性强，PAS算法能够自适应调节燃料电池运行功率，有着更好的综合性能。研究成果为全功率燃料电池汽车动力系统控制算法的设计奠定了基础。
In recent years, performance indicators such as power density, cold start at low temperature and durability of on-board Proton Exchange Membrane Fuel Cell (PEMFC) have been rising. The output power of Fuel Cell Engine (FCE) is increasing (up to 100~120 kW), and the powertrain configuration of Fuel Cell vehicle is gradually developing from extended range to full power type.Full-power fuel cell vehicles carry large fuel cell engine and battery capacity is small. Toyota's Mirai, launched in 2015, is a prime example. In this power configuration, the fuel cell engine needs to meet most of the power requirements of the vehicle. In the hydrogen-air-hydrothermal subsystems of fuel cell, due to the high power consumption, slow response time of the air system, and the large diffusion resistance of oxygen at the cathode side of the stack, local gas shortage is easy to occur in the dynamic process of power variation in a wide range of fuel cell engine. Therefore, it is necessary to design reasonable power system coordination control strategy to ensure the vehicle's dynamic performance, economy and durability.This paper focuses on a predictive hierarchical coordinated control strategy for full power fuel cell vehicles. The Predictive Adaptive softrun (PAS) energy management strategy and the decoupling control strategy of the air system are the two components of this strategy. Its basic idea is: through the LSTM algorithm to predict the a certain period of future time of average power demand, based on the average power coordinated control air compressor system and the back pressure valve, make the fuel cell engine output can maintain a certain degree of stability, but also with the vehicle power demand dynamic adaptive to adjust its power output.Firstly, a hierarchical model of the fuel cell vehicle powertrain is established, especially considering the dynamic characteristics of the FCE air system. Secondly, the double-input and double-output control algorithm of air system is designed and its validity is verified. Thirdly, a predictive adaptive softrun (PAS) strategy based on LSTM algorithm is designed, and three key parameters in the strategy are optimized. Fourthly, PAS algorithm, PF (Powerfollow) algorithm and SR (Softrun) algorithm are compared based on the same air system control strategy under NEDC operating condition and actual operating condition (including Beijing-Yanqing and Beijing-Zhangjiakou sections). The results show that there is a kind of SR algorithm is better under NEDC conditions because of the small vehicle average demand and good condition repeatability. In actual operating conditions, due to the large average power demand of the vehicle and the strong randomness of the working conditions, the PAS algorithm can adjust the running power of the fuel cell adaptively and has a better comprehensive performance. The research results lay a foundation for the design of power system control algorithm of full power fuel cell vehicle.
