基于GSA-PELM的锅炉NO_x预测模型

Optimization for NO_x Prediction Model from Boilers Based on GSA-PELM

为了更加精确地实现对电厂循环流化床锅炉NOx排放量进行预测,提出了一类基于并行极端学习机的GSA-PELM模型。由于PELM的泛化能力及精度依赖于其权值的选择,因而利用万有引力算法优化PELM的权值,采用从某火电厂300 MW的循环流化床锅炉在不同工况下实时采集的数据来检验模型的预测性能,并将GSAPELM模型分别与PELM模型、ELM模型、万有引力算法优化的最小二乘支持向量机模型(GSA-LSSVM)、GSA-ELM模型进行比较,仿真结果表明GSA-PELM模型的精度相比其它所有模型提高了9个数量级以上,可以更加有效、准确地用于预测火电厂锅炉的NOx排放浓度。

As to precisely predict the NOx emission from boiler,an integrated modeling method was establisbcd based on gravitation search algorithm （GSA） and parallel extreme learning machine （PELM） ,and used the model to forecast the power plant boiler NOx emission concentration. As the generalization ability and accuracy of PELM rely on the choice of weight,using the gravitation search algorithm optimization PELM weights. Test sample data were collected in a coal-fired power plant 300 MW circulating fluidized bed boiler under different conditions, the simulation consequences showed that compared with PELM model, ELM model, GSA-LSSVM and GSA-ELM model, GSA-PELM model the accuracy could be improved by more than 9 orders of magnitude and can be used more effectively and accurately to predict NOx emission concentration of boilers in thermal power plants.