300MW火电机组神经网络预测控制策略  

Predictive Control Strategy with Neural Networks for 300MW Fossil Fired Power Sets

作　　者：王世虎[1] 沈炯[1] 李益国[1] 

机构地区：[1]东南大学能源与环境学院,江苏南京210096

出　　处：《发电设备》2007年第6期435-439,共5页Power Equipment

摘　　要：针对在线滚动优化计算量大,间接算法梯度信息难以求取的问题,将神经网络多步预测控制应用于火电机组多参数协调控制中。提出了基于中心偏差商的优化算法,利用多元函数的偏差商替代函数的导数构成Hessen矩阵,从而获得优化算法的搜索梯度信息。通过对某300 MW单元机组双输入、双输出系统的仿真,验证了本算法的有效性。Because a lot of calculation work is required by on-line receding horizon optimization, which makes it rather difficult to acquire graded information by indirect calculation methods, an optimization method, based on multi-variable partial differential quotients （PDQ） to replace differentials of functions, for forming Hessen matrices, is proposed, wherewith detecting graded informations for optimization calculation are obtained. The effectiveness of the proposed calculation method has been vindicated by the simulation of a certain 300 MW block unit＇ s dual input and dual output system.

关 键 词：能源与动力工程 火电机组 神经网络 预测控制 多变量系统 优化 

分 类 号：TK323[动力工程及工程热物理—热能工程]