基于参数优化支持向量机的煤泥输送管道压力预测  

Pressure Prediction of Slime Transportation Pipeline Based on Parameter Optimized Support Vector Machine

作　　者：杨学存[1] 侯媛彬[1] 洪卫林[2] 王鑫照[1] 

机构地区：[1]西安科技大学电气与控制工程学院,陕西西安710054 [2]黄陵煤矸石热电有限公司,陕西黄陵727307

出　　处：《煤炭工程》2013年第12期112-115,共4页Coal Engineering

基　　金：陕西省自然科学基金资助项目(2009JM8002);陕西省教育厅专项资助项目(09JK712)

摘　　要：针对煤矸石热电厂煤泥输送管道堵塞问题,通过对实际现场的分析,确定煤泥输送系统浓料泵主缸压力预测是进行管道堵塞预测的必要前提,提出了基于网格法的支持向量机(GSVM)的浓料泵主缸压力预测模型,将网格搜索法用于支持向量机的参数优化。仿真结果表明,该GSVM预测模型参数寻优时间为6.55s,预测模型稳定,且相对误差在3%以内,能满足实际工程要求。与基于遗传算法的支持向量机(GA-SVM)预测模型相比较,GSVM预测模型在寻优时间和搜索稳定性方面优于GA-SVM预测模型,能够应用于煤泥输送系统实时压力预测系统。According to the blockage problems of the slime transportation pipeline applied in the coal rejects - fired thermal and electric power plant, with the actual site analysis, the paper defined that the pressure prediction of the master cylinder in the thick material pump of the slime transportation system would be the necessary premise to predict the pipeline blockage. The paper provided the pressure prediction model of the master cylinder in the thick material pump of the support vector machine based on the grid method and the grid search method was applied to the parameter optimization of the support vector machine. The simulation results showed that the parameter search optimized time of the support vector machine prediction model was 6. 55s. The prediction model was stable and the relative error was within 3% and could meet the actual engineering requirements. In comparison with the prediction model of the support vector machine based on the genetic algorithm, the GSVM prediction model would be better than the GA - SVM prediction model in the optimization time search and the stability search and could be applied to the real time pressure prediction system of the slime transportation system.

关 键 词：煤泥输送管道 堵塞 参数优化 支持向量机 网格搜索法 

分 类 号：TM614[电气工程—电力系统及自动化]