论文部分内容阅读
针对热耗率影响因素众多且呈现高度多重相关的特征,提出了偏最小二乘(Partial Least Square PLS)算法建立热耗率回归分析模型。在数据预处理方面对机组热耗率的主要因素和主要参数做了相关性分析,进而更合理地确定了偏最小二乘回归分析的数据表,有效建立了热耗率预测模型。预测模型的检验方式采用交叉有效性检验,选定对模型有显著改善的PLS主成分个数。通过实例验证了偏最小二乘方法能够有效解决自变量集合高度相关的问题。
Aiming at the characteristics of many and influential factors of heat rate, a Partial Least Square PLS algorithm was proposed to establish the regression model of heat rate. In the aspect of data preprocessing, the main factors and main parameters of unit heat rate are analyzed, and the data table of partial least-squares regression analysis is more reasonable. The heat rate prediction model is established effectively. The test of the prediction model adopts the cross-validation test, and selects the number of PLS components that have significantly improved the model. It is verified by examples that partial least squares method can effectively solve the problem of highly correlated set of independent variables.