基于故障预测与健康管理的固体氧化物燃料电池(SOFC)多约束高效率控制研究

Solid oxide fuel cell (SOFC) system is a clean and efficient power generation device, and its large-scale industrial application is of great significance to alleviate the energy crisis and protect the ecological environment. The performance degradation or failure of the system leads to performance decrease or even failure of the control algorithm, which is the bottleneck issue for the high efficiency and long life of SOFC system. It is the key technology for SOFC application that how to optimize the system controller parameters by fault prediction and health management methods. To this end, a multi-mode system incorporating system degradation and fault mechanism is constructed based on the prototype system. The migration rules of dynamic and steady-state characteristics and optimal operation region under different degradation rate are analysis by means of ‘system thermos characteristic and efficiency optimization’, and then health evaluation and management system are established. Design an improved nonlinear SOFC stack temperature observer based on ‘minimum dimension state space’ law and explore fast observer reconstruction method for key parameter observation. And then a ‘multi-step Gaussian Process Regression machine learning model’ is established for fault prediction and control strategy/parameters optimization, ensuring the performance of the system controller that designed with ‘TS fuzzy model based generalized predictive control algorithm’. Finally, the system can achieve ‘temperature constraints safe and system efficiency optimization during fast power tracking’ with the consideration of system degradation and fault, laying a solid theoretical foundation for the long-life operation of SOFC power system.

固体氧化物燃料电池(SOFC)系统是清洁、高效的发电装置，其规模化应用对缓解能源危机、保护生态环境具有重大意义。系统性能衰减或故障导致控制算法性能降低甚至失效是SOFC系统长寿命、高效率运行的瓶颈问题。利用故障预测与健康管理方法实时优化针对系统多约束热电耦合设计的控制器参数是其成功应用的关键。本项目以系统样机为依据，建立包含性能衰减与故障演变机制的机理模型，分析不同衰减下“温度安全及效率优化”的动静态特性与最优操作区间迁移规律，构建健康评价管理体系；设计改进“最小维状态空间”电堆温度观测器并探索快速重构方法进行定向观测；采用“多步高斯过程回归学习模型”预测故障，进而优化控制参数与策略，保障“基于数据驱动TS模糊广义预测控制架构”的控制系统性能，整体实现系统面向性能衰退与故障的“快速负载跟踪下热约束安全与效率优化”控制并验证，为SOFC独立发电系统的长寿命高效率运行提供理论基础。

固体氧化物燃料电池(SOFC)系统是清洁、高效的发电装置，其规模化应用对缓解能源危机 、保护生态环境具有重大意义。系统性能衰减或故障导致控制算法性能降低甚至失效是SOFC系统长寿命、高效率运行的瓶颈问题。利用故障预测与健康管理方法实时优化针对系统多约束热电耦合设计的控制器参数是其成功应用的关键。本项目首先以kW级SOFC系统样机为依据，系统梳理了SOFC衰减的机理因素，建立了包含性能衰减与故障演变机制的多模态机理模型；基于模型分析了不同衰减下“温度安全及效率优化”的动静态特性与最优操作区间迁移规律，获取了故障演化机制中的动态特征变化，构建了健康评价管理体系；为低成本准确获取电堆温度分布，进行热电协同管控，针对电堆温度分布估计，设计了改进“最小维状态空间”电堆温度观测器并探索了快速重构方法，基于机器智能算法对衰减条件下的工业非可测状态观测提出通用性方法；采用“多步高斯过程回归学习模型”进行了系统状态预测，并通过故障特征分析实现了故障预测；进而优化设计了的控制系统策略与参数，保障“基于数据驱动TS模糊广义预测控制架构”的控制系统性能，提高了系统效率约3%，有效避免了系统的故障发生，相比于原控制系统，延长了系统30%的服役时间，整体实现了系统面向性能衰退与故障的“快速负载跟踪下热约束安全与效率优化”控制。本工作为SOFC独立发电系统从衰减、故障演化机制，监控评价体系，温度分布估计及控制系统优化几方面进行了系统全面的研究，获得了系统面向应用的关键技术支撑，为长寿命高效率运行提供了理论基础及实现方案。