摘要
分析了制粉系统球磨机磨筒内负荷 (存煤量 )的各种影响因素及影响特性 ,提出了基于人工神经网络测量磨筒内负荷的软测量方法 ,给出了基于前向复合型神经网络的分工况学习的变结构式负荷测量模型及神经网络训练算法 ,正常工况下采用延时神经网络 ,而接近堵磨工况时则采用回归神经网络 ,工况的一种模糊划分方法同时被给出。离线训练及计算机仿真结果证实了所提神经网络软测量方法的可行性 ,为球磨机制粉系统的优化运行和自动控制奠定了基础。对某DTM3 5 0 / 60 0型球磨机的实测数据分析进一步证实了所提方法的有效性 ,且在计算结果的指导下 ,提高该球磨机制粉系统的出力约达 6t/h 。
Coal load in ball tube of pulverized ball mill system is effected by many factors, this paper analyzes these factors and their relations. The authors proposed a soft sensing method of coal load measuring based on NN(neural network), and a varying structure NN measuring scheme is presented, the NN is based on composite forward neural network model, an effective NN training algorithm is given. Delay NN is used for normal mode and recurve NN is used for blocked up mode separately, a fuzzy partition method of system modes is also given. Computer simulation results demonstrate the efficiency of the soft sensing strategy proposed, which is very helpful to the automatic control and optimal operation of the ball mill pulverized system. Computation and analysis of the data of a DTM350/600 ball mill obtained from operation field show that the method given by authors is feasible. More 6 ton coal powder per hour is produced by the guide of soft sensing computation results.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2001年第12期97-99,104,共4页
Proceedings of the CSEE
关键词
锅炉
制粉系统
球磨机
神经元网络
负荷
软测量
ball mill pulverized system
neural networks
composite neural networks
coal load in ball tube
soft sensing