应用神经网络对精煤灰分含量进行实时预报

Real Time Prediction of Ash Content of Clean Coal Using Neural Network

针对具有不确定性、时变性和复杂非线性关系的跳汰选煤过程,提出了精煤产品灰分含量的新型实时多步预测方法.本文基于Jordan神经网络构造了具有多作用因素输入和灰分含量动态时间序列反馈的实时动态建模预测模型,提出了BP算法和TD法相结合的网络学习新算法.该方法比传统预测方法具有更好的收敛性和适应性.应用结果表明,预测命中率和预测精度较高.

A novel real-time and multi-step predicting scheme of ash content of clean coal is proposed based on time-variation, uncertainty and complicated nonlinear relations in jigging process. A real-time and dynamic predicting model based on Jordan neural network including input of influence factors and dynamic time sequence feedback of ash content of clean coal was established. A new learning algorithm is proposed by combining the temporal difference methods with BP algorithm, which has better astringency and adaptability than the traditional predictive methods. The results applied in industry indicate that the predictive precision is quite high.