基于内部行为辨识的可变结构电池系统诊断与控制研究

Despite extensive R&D effort, current BMS technology remains inadequate for reliable and effective monitoring, diagnosis, control, and management of battery systems in electric vehicle, renewable energy, and smart grid applications. The main technical obstacle stems from variations of battery characteristics during operation, battery aging, mixing of different battery types, measurement noises, unreliability of traditional voltage monitoring approach, and inability of current estimation methods in achieving individualized and real-time battery cell characterization. Taking an holistic viewpoint, this project introduces innovative methodologies and develops new prototypes for a real-time, adaptive, and intelligent diagnosis and management system for battery hybrid power sources system. Our methodologies integrate multi-scale battery models of electrochemistry to circuit levels, combined parameter identification and state estimation of individual cells, stochastic time-series analysis for monitoring and diagnosis, hierarchical management frameworks using discrete-event and hybrid systems. The new system will introduce critically needed capability of real-time tracking and adaptation into the BMS technology, and as such substantially enhance reliability, energy efficiency, battery life cycles, and safety of battery systems in diversified utilities. The project will be carried out collaboratively by an team with diversified and complementary expertise in battery chemistry and material, battery modeling and system identification, battery diagnosis and monitoring. The team will work closely with major auto companies (GM, Ford, Shanghai Automotive) and battery companies (A123, LG Chem) for product verification.

电池系统的故障诊断与控制已成为当今国际研究热点之一。本研究探索一种基于内部行为分析与辨识的可变结构电池系统新的诊断机制与控制方法。包括：建立电池电化学模型、等效电路模型等多尺度模型，深入揭示电池特性与行为，为系统辨识与诊断提供依据；基于多尺度电池模型，提出准确实时参数辨识和状态估计方法，解决非线性电池系统信息获取、模型动态更新问题；探索基于系统参数辨识/状态估计、离散事件系统理论相融合的诊断机制，解决动态时变电池系统在线故障与失效的诊断与控制问题。通过本研究，能够扩展和完善电池系统诊断与控制的理论体系，为设计可靠、高效的电池系统，实现电池系统智能诊断与控制提供新的技术支撑，对形成具有自主知识产权的电池系统具有较高的实用价值。

随着新能源电池技术的发展，电池系统的可靠性与安全性成为当今电池技术发展的研究热点。本项目针对可变结构电池系统的可靠性与安全型问题，进行了一种基于内部行为分析与辨识的可变结构电池系统新的诊断机制与诊断方法的研究。包括：建立了电池等效电路模型模型，揭示了电池特性与行为，为系统辨识与诊断提供了基础；基于电池模型，提出基于随机分析与概率计算理论的自适应滤波算法，研究了电池系统实时参数辨识和状态估计方法，并解决了非线性电池系统信息获取、模型动态更新问题；探索了基于系统参数辨识/状态估计、离散事件系统理论相融合的故障诊断机制，尤其是基于DES（离散事件系统）理论，对电池系统的发热故障及老化问题进行了研究，结果表明，对于多模块可变结构复杂电池系统，基于DES的故障诊断方法可根据已知与未知参数的关联建模与计算，有效的解决复杂动态时变电池系统在线故障与失效的诊断与控制问题。通过本研究，扩展和完善了电池系统诊断与控制的理论体系，可为设计可靠、高效的电池系统，实现电池系统智能诊断与控制提供新的技术支撑，对形成具有自主知识产权的电池系统具有较高的理论指导意义和实用价值。