Application of support vector regression algorithm optimized by gradient descent method for analysing efficiency of boiler

Boiler efficiency is an important index to reflect the economy of boiler operation. It is of great practical significance to improve the safety and reliability of the unit. In order to make full use of the rich data provided by sensors and to reveal the correlation between them, this paper proposes a method based on the gradient descent (GD) method and the support vector regression (SVR) algorithm. It can realize the dimension reduction of boiler parameters and reveal the mapping relationship between several parameters and boiler efficiency. GD can roughly dig out the linear relationship between parameters such as the exhaust temperature of the high-pressure cylinder, the degree of undercooling of the condensate, the degree of vacuum of the condenser and the boiler efficiency. Meanwhile, the weight of each parameter can be obtained. Then, set a threshold to reduce the number of parameters and use them to train SVR model. The results of the case show that the method can realize the accurate fitting of the boiler efficiency under different working conditions. Therefore, this model can be used to optimize parameters' values and predict the efficiency of boiler.