标题:
基于FSVM锅炉烟气含氧量软测量Boiler Flue Gas Oxygen Content Soft Sensor Based on FSVM
作者:
刘真, 周玉国, 谢世龙
关键字:
FSVM, 烟气含氧量, 软测量, 预测模型FSVM, Flue Gas Oxygen Content, Soft Measurement, Prediction Model
期刊名称:
《Modeling and Simulation》, Vol.5 No.4, 2016-11-30
摘要:
针对电站锅炉燃烧过程存在的高度的复杂性和非线性问题，本文采用模糊支持向量机(FSVM)建立含氧量预测模型，预测在不同燃料量、总风量、总给水量等因素的影响下烟气含氧量的含量。选取模糊C均值算法(FCM)作为隶属度函数的设计方法，然后选取径向基核函数(RBF)和ε-SVR模型结构，其中惩罚因子和松弛变量的最佳参数值要用交叉验证法来选取。Matlab仿真实验结果表明，该方法有效地缩短了训练时间，提高了预测精度和模型的抗噪性，其性能优于一般支持向量机预测模型。
For power plant boiler combustion process of a high degree of complexity and nonlinear problem, fuzzy support vector machine (FSVM) is adopted to establish the prediction models of oxygen forecast in different fuel quantity, total air volume and total yield, total steam flow under the in-fluence of factors such as flue gas oxygen content. We select fuzzy c-means algorithm (FCM) as a design method of membership function, then select the radial basis kernel function (RBF) and ε-SVR model structure, and choose the penalty factor and the optimum parameter value of slack variable to use cross validation method. Matlab simulation experiment results show that this method can effectively shorten the training time, improve the prediction precision and the model of noise resistance; and its performance is superior to the general support vector machine forecasting model.