In order to realize the cold end optimization of direct air-cooled units

a cold end operation optimization method of air-cooled units is proposed based on the historical operation data of units and combined with data mining and deep learning algorithm. Firstly

the obtained historical operation data are screened in steady state and divided into working conditions. Combined with the Gaussian mixture model algorithm

the back pressure reference interval of the unit under multiple working conditions is determined. Then

the Spearman coefficient method is used to select the characteristic variables

and the back pressure prediction model of the direct air cooling unit is constructed in combination with the gated circulation unit. The back pressure optimization suggestions and early warning information are given by comparing the back pressure reference interval with the back pressure prediction value.Finally

the method is applied to a subcritical 300 MW air-cooled condensing steam unit. The results show that the back pressure optimization method proposed in this paper can give effective back pressure early warning information and realize optimal operation of cold end of the air-cooled unit.