基于模糊支持向量机的空气预热器温度预测

Temperature prediction of air pre-heater based on fuzzy support vector regression

空气预热器的温度检测对于火力发电机组的安全运行非常重要,提出利用模糊支持向量机回归算法,通过空气预热器历史温度信号与相关联的锅炉负荷信号建立空气预热器温度回归预测模型,对转子内部温度进行软测量。利用一种改进的模糊C均值聚类方法确定训练数据的样本权重,联立遗传算法与K折交叉验证方法优化模型的不敏感损失区域、惩罚因子和径向基核函数作用范围。通过某电厂600 MW机组的实验数据测试,提出的方法具有较高的准确度和泛化能力,较好的满足了提前检测空气预热器二次燃烧征兆的实际需求。

In this paper,Fuzzy support vector regression（ FSVR） is proposed to set up a regression model of air pre-heater temperature. The input and the output of the model are history temperature signals and the boiler load signals. The parameters of FSVR are optimized by Clustering algorithm and genetic algorithm. Fuzzy membership ship of the data is determined by an improved fuzzy c-means clustering method which combined with subtractive clustering and FCM. The best parameters of FSVR（ including regularization parameter,ε-insensitivity area and the kernel parameter） are optimized by genetic algorithm and K-fold cross-validation simultaneously. The simulation result shows that FSVR has higher accuracy and generalization than other ways for air pre-heater temperature model.This way can accurately predict the future temperature of air pre-heater and meet the demand of detecting the signs of secondary combustion.