基于Lasso-RF模型的燃气轮机启动过程建模分析

MODELING ANALYSIS OF GAS TURBINE START-UP PROCESS BASED ON LASSO-RF MODEL

建立燃气轮机启动过程的模型对指导启动操作以及优化启动时间具有重要意义。提出一种基于最小绝对收缩和选择算法(Least absolute shrinkage and selection operator,Lasso)的随机森林(Random Forst,RF)预测模型(Lasso-RF)。以三菱M701F4型燃气轮机历史运行数据为例,训练改进预测模型并对模型的准确性进行验证。结果表明,Lasso-RF模型的预测准确度、拟合程度都明显优于支持向量机(Support Vector Machine,SVM)模型。Lasso-RF模型不仅能高效精准地选择出燃气轮机启动过程模型的关键属性,并且能够对关键参数进行有效预测,对燃气轮机的启动过程优化具有一定的指导价值。

Establishing a model of the gas turbine start-up process is of great significance for guiding the start-up operation and optimizing the start-up time.Therefore,a random forest(RF)prediction model based on the least absolute shrinkage and selection operator(Lasso)is proposed,which is named Lasso-RF.Taking the historical operation data of Mitsubishi M701F4 gas turbine as an example,an improved prediction model was trained and the accuracy of the model was verified.The results show that the prediction accuracy and fitting degree of the Lasso-RF model are significantly better than the support vector machine(SVM)model.The Lasso-RF model not only can efficiently and accurately select the key attributes of the gas turbine start-up process model,but also can effectively predict the key parameters,and has certain guiding value for the optimization of the gas turbine start-up process.