基于灰色关联分析法及GSA-LSSVM的汽轮机排汽焓预测模型

摘要

The steam exhaust enthalpy is an essential parameter for the thermal economic diagnosis of steam turbine generator group. We determine the input variables of the model by steam turbine power equation and grey correlation analysis ( GCA) method, and optimize the punishment factor μ and nuclear radial range σ of least square support vector machine ( LSSVM) by gravitation search algorithm ( GSA) . The RBF kernel is selected as the kernel function of LSSVM through the comparative analysis. Based on the GCA-GSA-LSSVM, this paper establishes the mathematical model to predict the exhaust enthalpy of steam turbine, compares it with the BP neural network and RBF neural network, and analyzes its robustness. The results show that the prediction model of steam turbine exhaust enthalpy based on GCA-GSA-LSSVM has the advantages of high precision, strong generalization ability and strong robustness. This method provides a powerful tool for accurately predicting the energy saving potential of the unit.%排汽焓是汽轮发电机组热经济性诊断必不可少的一个参数.通过汽轮机功率方程与灰色关联分析(grey correlation analysis,GCA)理论确定了模型的输入变量,利用万有引力搜索算法(gravitational search algorithm,GSA)优化了最小二乘支持向量机(least squares support vector machine,LSSVM)的惩罚因子μ以及核径向范围σ2个参数.通过比较分析,选用RBF kernel为LSSVM的核函数.以GCA-GSA-LSSVM为基础,建立了预测汽轮机排汽焓的数学模型,并将其与BP神经网络、RBF神经网络进行对比,同时分析了该数学模型的鲁棒性.结果表明基于GCA-GSA-LSSVM的汽轮机排汽焓预测模型具有精度高、泛化能力强、鲁棒性强等优点,该方法为精确预测机组节能潜力提供了一种有力的工具.