Cold-end system operation optimization method based on clustering algorithm

The invention discloses a cold end system operation optimization method based on a clustering algorithm. The method comprises the following steps: acquiring historical data of unit operation measuring points and control variables; cleaning the historical data; extracting characteristic values of the cleaned data; carrying out characteristic value combination calculation to obtain new characteristic value measuring points meeting requirements; designing an optimal vacuum model, a condenser end difference model and a circulating water pump operation state model through mechanism analysis; performing unsupervised clustering feature tree modeling on the model state quantity to realize identification and classification of the operation condition; and performing comprehensive calculation on the energy according to the business connotation to obtain a real-time score of each model, namely a model result. The method depends on long-term real historical data, and the model credibility is high; the misalignment conditi