人工智能在发动机控制开发中的应用及前景

电动化、智能化、网联化、共享化 (CASE) 是汽车技术发展的趋势。根据国际能源署 (IEA) 的预测,到2050年,以包括插电式混合动力在内的电动化汽车将占有市场产品97%的份额。越来越严苛并要求通过实际行驶条件下检测的排放法规也对先进发动机技术,特别是发动机控制技术提出了新挑战。该文围绕基于模型的发动机控制开发中的优化问题,从前馈控制优化、反馈控制优化和动力总成全局优化3个层面,分析了人工智能 (AI) 技术在上述开发场景中的应用案例,展望了人工智能在发动机控制开发中的前景。研究表明,人工智能技术能将推动发动机控制开发中的3个融合：一是以发动机数字孪生技术为代表的信息系统与物理系统融合;二是以发动机多场景智能优化技术为依托的机器学习系统与经典控制系统融合;三是以动力总成域控制技术为基础的多源系统信息融合。这些融合将推动发动机技术的进一步发展,助力实现近零碳排放。

Connection, automation, sharing and electrification (CASE) are the future of vehicle and mobility systems. International Energy Agency (IEA) predicts that electric vehicles including plug-in hybrid vehicles will count for 97% of the market by 2050. The increasingly stringent emission legislations for CO₂ reduction, especially when involving real driving emissions (RDE) testing are the main challenges to the control system of the vehicle powertrains. This paper focus on artificial intelligence (AI) technologies for engine control developments that follow the standard model-based routine. By reviewing state-of-the-art AI technologies for feedforward control, feedback control, and global optimization at system level, the advantage and disadvantage of the AI technologies are compared and summarized. An outlook is provided based on the literature survey. It indicates that AI will promote the fusion of technologies in three representative domains, 1) fusion of cyber systems and physical systems, e.g., digital twin of engine; 2) fusion of machine learning systems and classical control systems, e.g., AI-based calibration of engine controllers; and 3) fusion of information from multiple sources, e.g., powertrain domain control network. The technology fusions in these three domains are expected to promote the development of advanced engines which aims to achieve zero emissions.