基于改进极根学习机的回转窑煅烧带温度预测方法

Forecasting Method of Rotary Kiln Calcination Temperature Based on Improved ELM

摘要:

针对传统算法预测回转窑煅烧带温度存在精度低、速度慢的问题,提出了基于改进极限学习机(ELM)的回转窑煅烧带温度预测方法;对ELM输入权值矩阵定义了变换系数,采用黄金分割法在给定区间内搜寻变换系数的最佳值,改进了ELM网络参数的确定方式,弥补了随机确定输入权值并且不作调整的缺陷,在保证ELM训练速度的前提下提高预测精度、减小模型随机性;实验结果表明,改进的ELM预测精度高、训练速度快、模型性能优,可满足工况恶劣的回转窑的生产需要。

Abstract:

In view of problems of low test precision and slow training speed in rotary kiln calcination temperature forecasting model with traditional algorithm, we proposed a forecasting method of rotary kiln calcination temperature based on improved extreme learning machine (ELM). A variable of the input weights’ transform coefficient of ELM was defined in the method, and golden section method was used to search the best value of the transform coefficient within a given range. The method improved the way to determine the network parameter of ELM, compensated the defects of input weights’ random generation and no adjustment, on the premise of ELM’s training speed, it increased forecasting accuracy and reduced the randomness of the model. Simulation result showed that the improved ELM model was superior to the original one, and had high accuracy of prediction and fast training speed, it can meet the need of rotary kiln in the bad working environments.