Research On Gas Turbine Combustion Optimization Based On CFD And Artificial Intelligence Algorithm

With the increasing requirements of environmental protection,China’s power generation industry has undergone profound industrial upgrades and adjustments.The coal-fired power generation industry is subject to many restrictions,and clean and efficient natural gas power generation has gradually attracted everyone’s attention.However,most of the core manufacturing and control technologies of gas turbine are controlled by the foreign gas turbine giants,while domestic gas turbine power generation enterprises bear the heavy responsibility of daily operation.The problems of excessive NO_x emissions,unstable combustion,and high cost pose great challenges for the safe and efficient operation of gas turbine combustion systems.Therefore,it is imminent to establish an efficient and systematic combustion optimization method.In this paper,an in-service 300 MW heavy-duty gas turbine is used as the research object,combined with CFD numerical simulation and AI intelligent algorithm,to deeply study the combustion characteristics of the gas turbine under variable conditions,and to simulate and summarize the data deviated from the stable operating conditions of the power plant,then combined with the actual operating data of the power plant to establish a new type of database,and build a gas turbine combustion optimization method based on the database,and create optimization software that can be used to guide the operation of the power plant in real time.To optimize the gas turbine combustion process,a full-size burner,including a swirl,was constructed.Steady-state gas flow calculations were performed using the Realizable K-method.The P-1 radiation model was used to describe the radiant heat transfer in the combustion chamber,the partial premix FGM combustion model was used to simulate the gas phase turbulent combustion,and GRI 3.0 methane-air chemical reaction mechanism was used in the calculation,including 53 components and 325 reactions,which could make the simulation results closer to the actual chemical reaction process.In order to prove that the mathematical model established in this paper can accurately reflect the actual flow and combustion process in the combustion chamber,the simulation results were compared with the data collected from the power plant for verification.The numerical error was all within5%,and the combustion characteristics of the gas turbine were further explored.With the results of numerical simulation as the theoretical basis and data supplement,the actual power plant data and the CFD numerical simulation data are combined to form a new training database,and a prediction model based on the LS-SVM algorithm is constructed,which can accurately simulate the combustion chamber inlet parameters and NO_x emissions,The relationship between the maximum amplitude of pressure pulsation and combustion efficiency,the average relative errors of the final prediction results are0.213%,0.329% and 0.453%,respectively.After that,when the working condition to be optimized does not meet the optimal working condition set,the genetic-ant colony algorithm(GA-ACO)and LS-SVM are combined to reduce NO_x emissions and the maximum amplitude of pressure pulsation,and improve combustion efficiency as the final combustion optimization goal.The average improvement rate of final combustion efficiency is 0.436%,The optimized amount of NO_x emission and the maximum amplitude of pressure pulsation is 8%-20%.At the same time,this method can not only be used to deal with gas turbine combustion optimization problems,but also can solve other complex optimization problems under high-dimensional nonlinear conditions.In addition,in order to construct the combustion optimization adjustment software for real-time docking with the power plant’s Mark VIe system,this paper integrates the actual data of the power plant and the numerical simulation results,and uses Matlab(R2018b)programming and software packaging methods to develop the visualization software for gas turbine combustion optimization,and establishes an intuitive user interface.In this combustion optimization software,the information display before and after optimization of various working conditions to be optimized is realized,including working condition data,numerical calculation cloud diagram,FFT(fast Fourier change)calculation results and optimization adjustment opinions.Finally,in the calculation of real-time operating conditions of the docking power plant,the optimization amounts of NO_x emissions,combustion efficiency,and maximum pressure pulsation amplitude were 15.52%,0.709%,and 16.63%,respectively.The optimization direction of operation parameters conforms to the actual operation of the power plant,and the optimization goal is basically achieved.The software is based on off-line numerical simulation calculation results to fit the flow velocity distribution,temperature distribution,NO_x generation distribution cloud diagram and pressure pulsation spectrum diagram of specific working conditions,which further promotes the visualization of optimization results.The fitting results are in line with the optimization practice of power plants and have practical application value.