含余热回收的卡车能源系统中多能流耦合机制及人工智能调控研究

Trucks play a pivotal role in the modernization construction in our country and they consume 49.2% of the total oil consumed by automobiles, becoming one of the main CO2 sources. Therefore, it is very important to improve energy efficiency of trucks. The Internal combustion engine of the truck only uses about 40% of fuel energy, the remaining energy is mainly discharged by the exhaust and jacket water, so it is quite an important way of improving internal combustion engine efficiency by waste heat recovery. The waste heat recovery system and the vehicle are coupled closely during the actual operation. Their coupling mechanism is quite complex and there are still some unsolved key problems: at the input end of waste heat recovery system, it is difficult to match well with multiple heat sources of engine during the whole working condition and meet the requirement of real application at the same time. At the output end, the traditional energy coupling structure can not satisfy well with vehicle energy demand all the time. Besides, there is no efficient vehicle energy management strategy by artificial intelligence. Therefore, the energy saving potential of waste heat recovery system under the actual road conditions has not been fully developed. In order to solve these problems, this study proposes using transcritical CO2 mixture rankine cycle as engine waste heat recovery combining with the energy coupling structure of parallel hybrid electric vehicles, and the whole vehicle energy management strategy is developed based on the deep Q-network algorithm, so as to realize efficient coupling of multi-energy flows in the vehicle during the whole road conditions.

卡车在我国现代化建设中有着举足轻重的作用，同时消耗了汽车消耗石油总量的49.2%，是CO2的主要排放源之一，因此提高卡车能效至关重要。而卡车内燃机仅利用了约40%的燃料能量，其余能量主要通过烟气和缸套水余热等被散失，所以余热回收是提高内燃机能效的重要手段。余热回收系统与实际运行的整车各能流深度耦合，其调控机制复杂且仍存在着关键问题，即在余热回收系统输入端传统系统很难与多余热源实时高效匹配并满足卡车上实际应用的要求，在输出端缺乏可实时响应整车能量需求的耦合模式，也缺乏含余热回收的整车智能能量管理策略，导致余热回收系统在实际卡车运行工况中的节能潜力还没有被完全开发。因此本项目提出采用CO2混合工质跨临界朗肯循环余热回收系统结合并联混合动力能量耦合模式，并采用基于深度强化学习算法的人工智能能量管理策略，最终实现整车能源系统全工况高效运行。

卡车消耗了我国汽车消耗石油总量的49.2%，因此提高卡车能效对减少碳排放至关重要。而卡车内燃机仅利用了约40%的燃料能量，其余能量通过烟气和缸套水余热等被散失，所以余热回收是提高内燃机能效的重要手段。余热回收系统与实际运行的整车各能流深度耦合且调控机制复杂，因此本项目目标是解决卡车运行工况下内燃机余热回收系统ORC(Organic Rankine Cycle)与整车能源系统协调耦合所面临的复杂能量管理问题。为此本项目从余热回收系统性能优化，与整车能量耦合的余热回收系统动态仿真控制，以及耦合余热回收系统的整车能量管理策略等方面开展了研究，并按预定计划完成了研究内容。. 通过对四大类典型多热源ORC余热回收系统进行充分优化分析，获得了它们各自与内燃机余热源特征的匹配机制，并制定了相应的余热回收系统构型选型方法；基于此方法针对不同内燃机开发了三款余热回收系统样机并作了大量性能测试实验，测试结果表明本项目所开发的样机最大可提升内燃机热效率3.7个百分点。进一步为研究余热回收系统动态特性和控制策略，本项目提出了高精度混合工质换热过程建模方法、标准的ORC余热回收系统动态仿真模型标定流程、机理骨架结合表征参数的高精度且高计算速度建模方法、以及基于深度强化学习的参数控制方法；据此建立了不同余热回收系统与整车耦合的高精度动态仿真模型，研究了余热回收系统动态特性并开发了控制策略，进一步确立了与整车通过混动耦合的技术路线。在此基础上，针对耦合余热回收的不同混动构型设计了基于规则、基于传统优化方法、以及基于深度强化学习的能量管理策略，并通过仿真证明了基于深度强化学习算法的能量管理策略的优异性。上述成果推动了余热回收技术的上车实际应用，目前负责人正在与中国重汽开展实车测试工作。