Application of deep learning modelling of the optimal operation conditions of auxiliary equipment of combined cycle gas turbine power station

Abstract

Reasonable condition adjustment of the auxiliary machinery can optimize the operation performance of the Combined Cycle Gas Turbine (CCGT) power station. In this paper an optimization method for the operation control of the auxiliary equipment based on deep learning is proposed and applied to an F-class gas turbine generation unit. A database of the optimal operating conditions is established to achieve the best economic benefit based on the analysis of two years of historical operation data. The control optimization models are built using the machine learning algorithms including the Multi-Layer Perceptron (MLP), Supporting Vector Machine (SVM), Gaussian process regression and linear regression methods and their accuracy has been compared. The input parameters of these models consist of the load rate, the heating output and the ambient temperature, with the output parameters including the electrical currents for the mechanical draft cooling tower, the high pressure feed pump, the medium pressure feed pump and the condensate pump. The MLP model is designed with different network hidden layer structures and provides the highest calculation accuracy with the least average error of 1.82 %. Different training data sizes are compared and the optimal control trajectory is analyzed. The result can provide useful references for the optimization of auxiliary equipment operations.