基于BP神经网络的循环介质密度控制系统设计

Design of circulating medium density control system based on BP neural network

以稳定循环介质密度值在所需范围为目标,运用MATLAB软件编程,建立了基于BP神经网络的循环介质密度控制系统。该系统以涡北选煤厂的生产数据为样本集,选择循环介质密度、磁性物含量和介质桶位为输入变量,调节阀开度、分流阀1开度和分流阀2开度为输出变量,经过训练、测试,确定实验数据隐层数为1、隐层节点为9,使误差基本稳定在0.002~0.012之间,符合控制系统的需求,从而为解决重介密度控制问题提供了一条思路。

In order to keep circlulating medium density stabilized within the required range, the BP neu- ral network-based density control system is designed through programming with MATLAB. The system circulating medium density, content of magnetic materials and medium level in tank as input variables, and with opening of regulator valve, opening of diverter valve 1 and opening of diverter valve 2 as output variables. It is determined through training and testing that the numbers of hidden layers and nodes of ex- perimental data are respectively 1 and 9 with an error basically stabilized in a range of 0. 002 ~ 0. 012, well up to the requirement. The development of the control system may serve as a reference for solving the problems in medium density control aspect.