基于小波SVM的电站锅炉飞灰含碳量预测

Prediction of the Unburned Carbon of Utility Boilers Based on Wavelet SVM

飞灰含碳量是影响锅炉热效率的一个重要因素,但由于电站锅炉机理复杂,很难建立能够用于飞灰含碳量实时预测和控制的机理模型。为此,将小波分析与支持向量机(SVM)算法相结合,提出基于小波SVM的飞灰含碳量预测模型。该方法可较好实现数据去噪和样本预处理,对因变量飞灰含碳量有较好预测能力,通过对样本自动筛选,实现预测模型自适应更新。最后,采用大型四角切圆燃煤锅炉热态实炉试验的运行数据对该算法进行了验证,并与神经网络预测模型进行了比较,结果显示提出的方法预测精度更高、效果更好。

It is very difficult to establish the accurate models on the unburned carbon content of utility boilers due to the complexities and time-variations of the boiler. An adaptive the unburned carbon in fly ash prediction algorithm for utility boilers was developed based on Wavelet methods and SVM. This algorithm not only can effectively perform the data preparing and samples pretreating, but also can abstract components from input factors. This paper also presents an adaptive SVM model that updates the sample space automatically. The predictive capacities of the SVM model were evaluated by the predicted the unburned carbon in fly ash of a high capacity boiler. Experiment results show that the wavelet SVM predictive model proposed by this paper has higher prediction accuracy by comparing with ANN models.