基于SVM的电站空冷凝汽器积灰特性建模

Ash fouling characteristics modeling based on support vector machine in air cooled condensers of power plant

电站空冷凝汽器(ACC)积灰严重降低了换热性能,建立有效的积灰特性模型是ACC换热性能在线监测与优化清洗的基础.以600MW机组参数为例,通过ACC传热过程理论计算模型,提取ACC积灰特征参数,使用支持向量机算法,利用现场实验获取的检测数据,建立ACC换热面积灰状态下换热系数预测模型,进而推导ACC清洁系数,从而实时监测积灰清洗状况.结果表明:所建模型可实现换热性能的快速监测,有效降低换热系数的计算复杂度;模型验证样本均方根误差、绝对百分比误差和决定系数分别为0.60、2.13%和0.9838,预测性能良好.该模型可为开发大型空冷机组ACC积灰监测与智能清洗系统提供依据.

The ash fouling seriously reduced the heat transfer performance of air cooled condensers(ACC),and establishing an effective characteristics modeling of ash fouling is the basis to online monitoring and optimization cleaning.Based on a 600MW power plant,the fouling characteristics parameters are extracted by the structure and heat transfer process,and a heat transfer coefficient prediction model is established under fouling accumulation condition by the test data obtained from site experiment using support vector machine(SVM).Furthermore,the clean coefficient of ACC is calculated and the fouling clean condition can be monitored accurately.The results show that the established modeling can quickly monitor the heat transfer performance and reduces the calculation complexity of heat transfer performance.The value of root mean squared errors(RMSE),mean absolute percentage errors(MAPE)and square correlation coefficient(R^(2))of the SVM model for the verification data are:RMSE=0.60,MAPE=2.13%and R^(2)=0.9838.The prediction model and the method can provide a basis for developing online monitoring and intellective cleaning system of ACC plant.