基于支持向量回归的凝汽器清洁系数时间序列预测

分析了凝汽器水侧污垢形成的机理，得到凝汽器清洁系数随时间变化的基本规律。提出采用支持向量回归时间序列预测法来预测凝汽器清洁系数。简要介绍了支持向量回归的理论基础，建立了凝汽器清洁系数时间序列预测模型，利用某300MW机组的数据，对模型进行了校验，探讨了参数的选择。并同径向基函数神经网络预测模型进行比较，结果表明，支持向量回归模型在预测性能方面明显优于RBF神经网络方法，并且模型具有较好的预测精度和泛化能力，为凝汽器真空降低故障的诊断，奠定了一定的基础。

Time Series Prediction of Condenser Cleanness Coefficient Based on Support Vector Regression

The analysis is done on mechanism of foul-forming firstly. Basic rule of cleanness factor changing with time is taken. Cleanness coefficient of the condenser is predicted based on support vector regression(SVR) time series prediction method. SVR is introduced simply. Time series prediction model of the condenser based on SVR is built, and taking a 300MW generation unit as an example, the model is proved to be rational and how the parameters are elected is discussed. At last, the calculation result based on SVR is compared with the result based on radial basis function. As a result, it is shown that the prediction model based on SVR is exceeds the prediction model based on RBF and the prediction models based on SVR has powerful prediction precision and generalization. The prediction model based on SVR lays the foundation on failure diagnosis of condenser vacuum.