基于模型与数据驱动的固体氧化物燃料电池系统热电耦合特性分析与管控研究

"Thermoelectric coupling characteristics dynamic analysis and control for solid oxide fuel cell (SOFC) system” is one of the key techniques to achieve SOFC stable and high conversion efficiency operation. Without integrating fully mechanism model and the data processing model, it is difficult for traditional system model architecture and dynamic analysis control strategy to guarantee dynamic process meeting the SOFC thermal constraints and achieve high efficiency of output power without fuel starvation. Therefore, based on the innovation of mechanism and technology feasibility of system structure, a thermocouple mechanism model of modified SOFC system is established, which is combined with data driven to make the dynamic and static analysis fulfill the following requirements, including system thermal characteristic safety constraint, fuel starvation avoidance and output power efficiency optimization. The effective working range of the modified SOFC system external adjustment parameters with different demand power is obtained in the open-loop state. And the influence mechanism between the electrical efficiency optimization and the adjustment parameters are revealed. Then the optimized reference trajectory is proposed for the SOFC system thermoelectric coupling control. Now that sealed stack structure makes it difficult to measure the stack internal temperature directly, based on the "minimum dimensional state space optimized and combination of measurable SOFC thermoelectric parameters", the stack internal temperature observer is designed, which can feedback real-time stack internal temperature from the thermoelectric characteristic parameter with constraint combined with mechanism model analysis and experimental data. Finally, a constrained model predictive control strategy is established base on Bayesian learning for SOFC system, which can be used to realize fast steady-state and high-efficiency operation of the external load demand power for SOFC system, and can also offer a theoretical basis for the optimal design of independent SOFC generation system.

“固体氧化物燃料电池（SOFC）系统热电固有特性分析与管控”是实现SOFC稳定高效率运行的关键环节。传统系统模型架构及分析管控方法在机理与实验模型之间未充分融合，难以保障“SOFC热特性与避免燃料亏空”约束下高效率按需输出功率。为此，基于机理与结构可实现性创新，搭建新型SOFC系统机理模型，并与数据驱动相结合，面向“系统热安全约束、避免燃料亏空及电效率优化”进行动静态分析，获取开环状态下不同负载调节参量组合的优化区间，揭示电效率与调节参量间影响机理，提炼出面向系统热电协同管控的优化参考轨迹；同时，构造“最小维易测SOFC热电参变量组合”状态观测方法，设计难以直接测量的电堆内温度观测器，实时反馈“机理模型分析与实验数据驱动相结合”的热特性约束参量。最后，建立“基于贝叶斯学习的带约束模型预测控制”策略，可物理实现SOFC快速稳态高效率输出需求功率，为SOFC系统优化提供理论依据。

本项目完成了“固体氧化物燃料电池（SOFC）系统热电固有特性分析与管控”工作，实现SOFC稳定高效率运行。本项目扭转了传统系统模型架构及分析管控方法在机理与实验模型之间未充分融合，难以保障“SOFC热特性与避免燃料亏空”约束下高效率按需输出功率的局面。基于机理与结构可实现性展开了创新工作，搭建了新型SOFC系统机理模型，并与数据驱动相结合，面向“系统热安全约束、避免燃料亏空及电效率优化”进行了动静态分析，获取了开环状态下不同负载调节参量组合的优化区间，揭示了电效率与调节参量间影响机理，提炼出了面向系统热电协同管控的优化参考轨迹；同时，构造了“最小维易测SOFC热电参变量组合”状态观测方法，设计了难以直接测量的电堆内温度观测器，实现了“机理模型分析与实验数据驱动相结合”的热特性约束参量的实时反馈。最后，建立了“基于贝叶斯学习的带约束模型预测控制”策略，在物理层面实现SOFC快速稳态高效率输出需求功率，为SOFC系统优化提供了理论依据。