In order to solve the problem of comprehensive evaluation for multi-indicator system involved in operation state evaluation of thermal power units

the principal component analysis and cluster analysis are combined and applied to the comprehensive evaluation of power units. This method utilizes the principal component analysis to perform dimension-reduction

standardization and decorrelation for unit operation indicators

then the principal components affecting the unit operation state are extracted. Moreover

the extracted principal components are considered as a new data matrix for cluster analysis

which avoids the disadvantage of subjectivity of the classical clustering method when choosing the clustering variables. Finally

the principal component scores and the cluster classification are combined for the comprehensive evaluation. The case study result verified the effectiveness of the above method.`