Research On Combustion Stability Of Gas Turbine

Electric energy is indispensable to human life, in recent years, with the national energy structure adjustment and improvement of environmental protection requirements, gas turbine units have become an important direction of China’s energy structure adjustment. Compared with the traditional coal-fired units,gas turbine units are faster in the process of starting and stoping.It can content the requirement of dispatching automation system of power network and its flue gas discharge is more cleaner. The gas turbine to run safely and stably, we must overcome in the stage of start stop and load change stage of combustion instability situation. The gas turbine operation process if the combustion instability, will be on the gas turbine components especially the structural damage caused by internal combustion.In this paper, the research content and the main achievements are in these following respects:1. This paper introduces three factors which reflect the combustion stability:the acceleration signal,the pressure fluctuation and the exhaust temperature of gas turbine. And the three factors are selected to monitor the stability of gas turbine combustion state. Research shows that acceleration and pressure fluctuation are sensitive on the combustion state. Exhaust temperature distribution change radar chart also has an intuitive judgment on the combustion chamber flame distribution, so the three sets of parameters can be used to judge the combustion stability of the gas turbine.2. The combustion process of gas turbine is an extremely complex model, the units currently used in China are lack of research on combustion model, and lacking precise criterion to judge the combustion stability.The data of this paper is based on Beijing caoqiao power plant in recent years. We collected a lot of data and using the BP neural network algorithm to train repeatedly.It can predict combustion stability with great accuracy.3. This paper introduces the particle swarm optimization algorithm model, and elaborates its advantages and disadvantages. In order to improve its faults that trapped in local minima easily and unable to achieve global optimization of faults, the method which puts the modification method and the extremum disturbance method in the particle swarm optimization algorithm model. And then using improved particle swarm algorithm, optimize the weights and thresholds of the BP neural network, build based on BP neural network and improved particle swarm model, enhance the training accuracy of the neural network model.4.The data this paper used is based on Beijing caoqiao power plant.We take values of the acceleration signal and the pressure fluctuation signal into improved PSO-BP neural network algorithm for the modeling and simulation and compare with the classical BP neural network results,and then we discuss the results of them.