基于神经网络解耦线性化方法的锅炉主汽压力控制  

THE CONTROL OF BOILER PRESSURE BASED ON NEURAL NETWORKS DECOUPLING LINEARIZATION

作　　者：于达仁[1] 徐志强[1] 郭洪波[1] 范轶[1] 

机构地区：[1]哈尔滨工业大学,黑龙江哈尔滨150001

出　　处：《中国电机工程学报》2002年第5期143-147,共5页Proceedings of the CSEE

基　　金：国家自然科学基金项目(59776047);霍英东基金项目。

摘　　要：锅炉对象的动态特性受到汽轮机运行状况的很大影响,必须采取解耦措施才能正常运行,但是由于其动态特性具有很强的非线性、时变性和不确定性,常规的线性解耦设计很难获得较好的效果。另一方面锅炉每天的运行是重复的,从机组的运行数据中可以充分了解对象的动态特性。该文基于神经网络a阶逆系统方法提出了一种锅炉主汽压力非线性控制系统的设计方法,利用神经网络强大的自学习和自组织能力,发掘机组运行数据包含的对象动态 特性的信息,克服对象的不确定性的影响,实现了系统的大范围解耦线性化。仿真结果表明,此控制策略能有效改善系统在大范围内工作时的控制品质和自适应能力。The dynamic behaviors of the boiler are greatly affected by the operation of turbine because of the strong nonlinear cross coupling between boiler and turbine. The plant controller should be able to compensate this nonlinear cross coupling in control algorithms. The conventional fixed-gain controllers are not idea in these circumstances. This work analyses the dynamic mechanism of a boiler, and then describe an application of neural networks decoupling linearization method in the pressure control of a drum type boiler. This technique provides the possibility to make use of the plant dynamics information involved in the operating data, so that we can attenuate the affects of plant nonlinearities and uncer-tainties. Then by this method, the original nonlinear system can be linearized and decoupled over a wide operating range with-out the priori of the exact mathematic model of the system. The simulation results show that substantial improvement in control can be achieved over a wide range of operating conditions.

关 键 词：神经网络 解耦线性化 锅炉 主汽压力控制 

分 类 号：TK223.7[动力工程及工程热物理—动力机械及工程]