Control and machine learning for internal combustion engines and their exhaust aftertreatment systems

内燃机及其排气后处理系统的控制和机器学习

• 批准号: RGPIN-2022-03411
• 负责人: Koch, Charles
• 金额: $ 3.35万
• 依托单位: University of Alberta
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=760207
• 关键词: Control; machine; learning; internal; combustion

The short to medium term objective of this research program is to develop methods to reduce CO2 and other harmful emissions from internal combustion engines using carbon free or renewable fuels with the focus on heavy duty engines. Canada and many other countries, plan to reduce greenhouse gas emissions to zero by 2050. This is causing tremendous changes and innovation in the transportation sector. Heavy Duty Class 8 trucks in North America move 60% freight ton-km and produce over 75% of the CO2 emissions of road freight. One promising way to reduce CO2 emissions in these transport trucks is to decrease fuel usage by maximizing the thermal efficiency and using renewable fuels or zero carbon fuels such as hydrogen (H2). The impact of this work is to quickly reduce CO2 emissions from these trucks in Canada. For example, a H2/Diesel dual fuel 40:60% (energy split) in heavy duty freight trucks reduces the life cycle CO2 32% when using blue H2 for trucks in Alberta with a projected reduction of 1.5 megatonnes of CO2/year by 2039. Current Internal Combustion Engine calibration methods have become so complex there is strong industry demand to embed knowledge and constraints of the system in a model based control. Innovative engine concepts such as H2 or H2/Diesel dual fuel, which can quickly reduce CO2, are being delayed into the market due to the complexity and cost of current control/calibration methods to meet Real Driving Emission (RDE) requirements. The focus of this proposal will be the development of methods and their application of Machine Learning control to H2/Diesel dual fuel, 100% H2 fuel and other renewable fuels in heavy duty freight trucks to quickly reduce CO2. A detailed physical understanding of the combustion and exhaust aftertreatment system including the sensors/actuators is needed. This is incorporated into a simulation model suitable for model based control. For example, to decrease CO2 in freight trucks using Diesel/H2 or 100% H2 requires using an understanding of knock, preignition, NOx and particulates emissions to develop control strategies. The long term objective is that the proposed research will lead to ground breaking advances in (1) In-cycle control for engine combustion enabling advanced combustion; (2) new methods of combining Machine Learning with Model Predictive Control for complex constrained engineering systems; (3) Machine Learning-Control for robust and easy to calibrate engine control that minimizes fuel and emissions using carbon neutral fuels; and (4) training of highly qualified people to work in the H2 economy. This will have a long term impact to allow Canada to remain competitive in complex engineering technologies, such as transportation, that require control for optimal performance.

该研究计划的短期至中期目标是开发使用无碳或可再生燃料减少内燃机二氧化碳和其他有害排放的方法，重点是重型发动机。加拿大和许多其他国家计划到2050年将温室气体排放量减少到零。这在交通领域引起了巨大的变化和创新。在北美，重型8级卡车运输60%的货运吨公里，产生超过75%的公路货运二氧化碳排放量。减少这些运输卡车中的二氧化碳排放的一种有希望的方法是通过最大化热效率和使用可再生燃料或零碳燃料（如氢气（H2））来减少燃料使用。这项工作的影响是迅速减少加拿大这些卡车的二氧化碳排放量。例如，当阿尔伯塔的卡车使用蓝色H2时，重型货运卡车中的H2/柴油双燃料40：60%（能量分配）将生命周期CO2减少32%，预计到2039年每年减少150万吨CO2。 当前的内燃机校准方法已经变得如此复杂，存在将系统的知识和约束嵌入到基于模型的控制中的强烈工业需求。创新的发动机概念，如H2或H2/柴油双燃料，可以快速减少CO2，由于目前的控制/校准方法的复杂性和成本，以满足真实的驾驶排放（RDE）的要求，被推迟进入市场。该提案的重点将是开发方法及其在重型货运卡车中对H2/柴油双燃料、100% H2燃料和其他可再生燃料的机器学习控制应用，以快速减少CO2。 需要对燃烧和排气后处理系统（包括传感器/执行器）有详细的物理了解。这被合并到适合于基于模型的控制的仿真模型中。例如，为了减少使用柴油/H2或100%H2的货运卡车中的CO2，需要利用对爆震、早燃、NOx和颗粒排放的理解来开发控制策略。长期目标是，所提出的研究将在以下方面取得突破性进展：（1）发动机燃烧的循环控制，实现先进燃烧;（2）将机器学习与模型预测控制相结合的新方法，用于复杂的受限工程系统;（3）机器学习控制，用于鲁棒且易于校准的发动机控制，使用碳中性燃料最大限度地减少燃料和排放;以及（4）培训高素质的人在H2经济中工作。这将产生长期影响，使加拿大在复杂的工程技术方面保持竞争力，例如交通运输，这些技术需要控制以实现最佳性能。