Abstract: For the low economic performance caused by coal pulverizing system multi-variable, strong coupling and the inaccuracy material level measurement, a multi-neuron PID decoupling control of coal pulverizing system based on the material level’s of self optimizing is proposed. Firstly, the mill maximum output has been found with the real-time material level searching optimization algorithm according to the mill sound and mill load parameters, and the system can run at the best operating point. Then the weight improving neural network PID algorithm is used according to the material level controlling quantity at the best operating point. The simulations and results of actual control show that the algorithm has higher accuracy and is able to adapt to the stable control of exit’s temperature, inlet’s negative pressure and the material level under different conditions. Not only closed-loop automatic control of the milling system has been achieved, but also the efficiency of the pulverizing system has been improved, which makes the consumption of flour milling reduced, in order to achieve the optimization target of energy saving ultimately.

Key words: decoupling control, coal pulverizing system, multivariable, neural network, self optimizing

摘要： 针对制粉系统多变量、强耦合、料位测量不准确造成的经济性能低等特性，提出了基于料位自寻优多神经元PID制粉系统解耦控制。根据现场采集的磨音信号、磨机负荷等参数，采用料位自寻优算法能实时搜索磨机最大出力，使系统始终运行在最佳工作点；根据系统在最佳工作点的料位控制量采用改进权重的神经网络PID算法，实现各层神经元的连接权重值调整，降低系统控制误差，使系统实现解耦控制。仿真和实际控制结果表明，该算法具有较高的准确率，能适应不同工况下出口温度、入口负压以及料位的稳定控制，不仅实现了制粉系统的闭环自动控制，而且提高了系统制粉效率，降低了制粉单耗，最终达到节能目标。

关键词: 解耦控制, 制粉系统, 多变量, 神经网络, 自寻优