多模块核能系统的自愈控制方法研究

The multi-modular nuclear power plant (MNPP), which adopts the scheme of multiple small modular reactors (SMRs) driving a common secondary loop system, is strong in both the aspects of nuclear safety and economic competitiveness, and is a promising trend in the development of advanced nuclear energy systems. MNPP is not only a complex nuclear thermal fluid network (NTFN) including nuclear fission reaction, thermal transfer and fluid transportation but also a complex cyber-physical system composed of numerous process, measurement and control equipment. To guarantee the operating reliability for MNPPs, it is necessary to realize the resilience function against the anormalies caused by physical malfunctions and cyber-attacks. In this project, the minimal dynamic model of NTFN is proposed. Then, by the use of this minimal model, the analysis of the influence of anormalies to the dissipation, observability, controllability and identifiability is performed, based on which the conditions for both anormaly detectability and compensability are given. Moreover, by proposing the analytically redundant nonlinear observers, the anormaly detection method is built, and by strengthening the closed-loop dissipation, the anormaly compensating control approach is given, which finally results in the resilient control theory of MNPPs. The research results of this project are both scientifically and practically meaningful for solving the key bottleneck technical problem of developing MNPPs.

多模块核能系统采用多座小型模块式反应堆驱动公共二回路的方案，具有很强的核安全性和经济竞争力，是先进核能系统发展的主流趋势之一。多模块核能系统既是包含核裂变反应、传热和流体输送过程的复杂核热工流体网络，也是包含众多过程设备、测控设备的复杂信息物理系统，为了保证运行可靠性，需要对物理空间设备故障和信息空间攻击行为引起的异常现象具有自动检测与补偿功能。本项目在建立核热工流体网络最小动态模型的基础上，通过揭示异常现象对耗散性、能控性、能观性及参数和拓扑结构可辨识性的影响，提出异常可检测性与可补偿性的判别方法，通过建立非线性冗余观测器组，提出异常检测方法，通过增强闭环系统耗散性，提出异常补偿控制方法，最终形成多模块核能系统的自愈控制方法，并通过开展数值和半实物仿真实验来验证其有效性与可行性。项目研究成果将对突破多模块核能系统的关键技术瓶颈具有重要科学意义和应用价值。

多模块核能系统采用多座小型模块式反应堆驱动公共二回路的方案，具有很强的核安全性和经济竞争力，是先进核能系统发展的主流趋势之一。多模块核能系统既是包含核裂变反应、传热和流体输送过程的复杂核热工流体网络，也是包含众多过程设备、测控设备的复杂信息物理系统，为了保证运行可靠性，需要对物理空间设备故障和信息空间攻击行为引起的异常现象具有自动检测与补偿功能。本项目面向工程需求集中发展了多模块核能系统的自愈控制理论，包括：换热器网络无源性化控制方法、运行可靠性监测的高阶扩张状态观测方法、过程故障诊断的布尔网络参数识别方法、反应堆异构冗余执行器的容错控制方法、性能补偿控制的动态矩阵预测控制方法、神经网络预测控制方法和强化学习控制方法。为了深入开展控制方法性能验证，本项目完成了多模块高温气冷堆核电站控制系统半实物仿真平台集成和组态，开展了典型工况下的半实物仿真试验，为后续工程实施奠定了坚实基础。在项目执行阶段，发表SCI期刊论文16篇，其中影响因子大于7的有10篇，授权发明专利4项，培养博士后1人、博士生2人、硕士生2人，合作举办了第一届核电数字化博士生论坛。