Gas turbine is one of the most important large-scale equipment power plants at present, which has been widely used in the industrial field. Mechanical performance test is a very important parameter for gas turbine. Neural network algorithm can play a great role in system data processing and monitoring. Therefore, in order to observe the dynamic performance of gas turbine, neural network is used to study the data changes related to intermediate cooling and regeneration of gas turbine. This paper mainly uses the experimental method to compare, control the variable of pressure ratio, and analyze its dynamic performance through computer group efficiency, fuel consumption rate and the quality of carbon dioxide and oxygen. The experimental results show that, under rated load, the unit operates with variable pressure ratio. When the pressure ratio is less than 11, if efficiency is required, the unit power can be kept basically constant by changing the fuel quality.