实时优化与分布式控制集成算法

Integration of Real-time Optimization and distributed predictive control

摘要:

预测控制作为一种以预测模型为基础的先进控制算法，分布式预测控制广泛应用于复杂高维的复杂大系统控制中。提出了针对大规模过程的实时优化与分布式预测控制集成算法，包含稳态目标计算层和动态分布式控制层。在稳态目标计算层根据当前系统的运行状况进行集中优化，在系统运行的每一时刻计算出相对的全局最优值，并将其传递到下层进行控制。在动态控制层将复杂的大系统分为若干个相对独立的子系统，并且充分考虑各个子系统之间的关联和耦合，用分布式预测控制算法对上层计算得到的相对的全局最优值进行跟踪，提高系统在动态情况下的控制性能。仿真应用表明，此方法的优点在于保证全局最优的同时，降低了计算的复杂度，并且实现了经济目标。

Abstract:

Abstract：In order to get rid of the disadvantages that verdict information is neglected of traditional fault feature extraction method, Complete Kernel Fisher Discriminant (CKFD) feature extraction algorithm based on Chaos Cuckoo Search (CCS) and its application on fault diagnosis is presented. Apply the algorithm in the feature extraction of fault data of Fluid Catalytic Cracking Unit (FCCU). First the data are mapped to a high-dimensional space，and CKFD based on FPP model is used to divide transform domain into two spaces, then regular and irregular discriminate vectors are extracted from the two subspaces respectively by fisher criterion function, last the CCS algorithm is used to optimize the fusion coefficient of two vectors to obtain the optimal discriminant vector. Finally, the data are classified by the nearest-neighbor classifier based on Euclidean Distance to validate the validity of the arithmetic. It can be seen from the simulation that the presented algorithm can raise the accuracy of fault classification and it can also improve stability of multi-fault classification.