基于支持向量机的电站锅炉燃烧稳定性判别

以某600MW单元机组锅炉运行时的火焰信号为分析对象,利用支持向量机研究了锅炉燃烧稳定性判别问题。提取了火检信号的均值、方差、峰峰值和均匀度等4个特征量,然后比较了采用3种不同核函数的支持向量机对同一火检信号样本组的训练效果,其中模型参数通过交叉验证的方法确定。分析表明,径向基核支持向量机判别准确率最高。最后测试结果验证了支持向量机用于火焰燃烧稳定性判别,具有很好的分类和泛化能力。

Research on judging of combustion stability for utility boiler based on Support Vector Machines

Support Vector Machines was presented to judge the combustion stability of coal fired boiler based on the flame signals from the boiler of a 600MW boiler-turbine unit.Firstly the four features of the flame signal were extracted,which were average of the flame signal,variance of the flame signal,peak-peak value of the flame signal,rate of the sum of the power spectrum values in different frequency zones.Then comparing with the training results of the SVMs with three types of kernels,it shows that the SVM with radial basis function kernel function had more accurate than the others.To get better classification result,the model parameters of classifiers were tuned with cross-validation.Finally the test result suggests that SVM is effective for classification combustion flame and has better generalization ability.