大规模复杂系统通常由许多相互连接的子系统组成。集中控制结构难以满足系统内部计算复杂性、鲁棒性、可靠性以及通信带宽等要求。多级分布式模型预测控制是处理大规模复杂系统的有效手段。由于多级分布式模型预测控制需在局域控制器间传递信息以达到全局稳定性，因此相对于集中预测控制，其理论与应用更加复杂。针对多级分布式模型预测控制系统结构选择、分布式状态估计、以及保证闭环稳定性能的优化算法展开研究。现代电力系统是典型的地理分散的大规模系统，尤其是以风能为主的间歇性可再生能源发电系统介入时会引起电网波动。运用多级分布式模型预测控制实现电力系统自动发电控制（AGC），满足电力系统多种约束条件下，使AGC实现以减少机组疲劳和磨损为核心的经济目标。运用非线性系统模糊神经网络建模方法、启发式算法、遗传算法等去解非线性系统约束问题，实现多级分布式模型预测控制中的多目标优化，从而实现电力系统的经济运行。

Large scale systems are often composed by many interacting subsystems and can be difficult to control with a centralized control structure due to the required inherent computational complexity, due to robustness and reliability problems and due to communication bandwidth limitations. For all these reasons, distributed and hierarchical model predictive control have been developed and applied to be an effective way of copying with these large scale systems. The design of distributed MPC systems needs to transmit the information among local regulators to achieve global stability and performance results, thus is more complicated than centralized control. This research is devoted to selection of the control structure, the distributed state estimation, and the effective optimal algorithms with guaranteed properties and stability. Modern power system is the typical large-scale, geographically expansive systems. A distributed MPC framework is appealing in this context, especially when greater utilization of intermittent renewable resources, such as wind generation, brings with it power flow fluctuations. Most interconnected power systems rely on automatic generation control (AGC) for regulating system frequency and tie-line interchange. With distributed and hierarchical model predictive control, the economic objective is to minimize the number of AGC requests and unit generation reversals, which gradually create fatigue in generating units. Consequently, the operational and maintenance costs can be reduced. Neuro-fuzzy networks modelling, heuristic algorithms and genetic algorithms are utilized to solve the nonlinear constraint optimal problem, in order to realize the multi-object optimization in power system economic running.