Performance Model Of Large-scale Combined Cycle Units And Combustion Optimization Adjusted Analysis

The large-scale combined cycle generator is considered as a high-efficient environmentally friendly way of electricity generations. However, the natural gas generation poses a significant lower proportion in China electric power energy structure comparing to the developed countries. Until the end of 2009, the China total installed power capacity reached 874097.2MW, in which the gas-fired units installed power capacity was 24029.6MW with a percent of 2.75%. Predicted by experts, the natural gas electric generation proportion will reach 6%~7% till year 2020. The extensive exploitation, import and application of the natural gas in China provided solid foundation for the natural gas electric generation. Besides, demanded by the severe environment pollution situation and the government energy saving policy, the sustainable development of the natural gas electric generation will be supported continuously.In this thesis, a Mitsubishi M701F gas-fired turbine was chosen as the study case. The M701F gas-fired turbine was composed by 17-stage high-efficiency axial flow compressor, a combustion chamber with 20 burner engines distributed along the rings and a 4-stage reaction turbine set. The M701F gas-fired turbine is characterized by the duplex bearing, the uniaxial structure, the cold end electric driven and the axial exhaust, while the combined cycle is designed in uniaxial structure.The Neural Network is highly appreciated not only by its powerful nonlinear mapping but also the self-learning, self-organizing and self-adaptive capability, along with its identification and potential in nonlinear control systems. Without requiring a certain identification system pattern, neural network system conveys its nonlinear mapping capability in its neuron function and network structure with tunable parameters distributed in inner network connection weighting matrices. When in identification of a multi-input-multi-output system, the neural network only needs a practically satisfying knot number of the input and output layers. As for its simplicity without requiring complicated learning algorithm, neural network suits itself in predictive multi-input-multi-output system modeling.For a safe high-efficient and stable combustion, a multi-parameter coupled nonlinear system is applied to describe the combustion control system. The inputs of the combustion control system are determined by the gas-fired main control signal output (CSO), the natural gas processing system (natural gas parameter) and pressing machine control system (ie. IGV), while the system outputs determined the gas-fired turbine blade channel temperature, the combustion pressure pulsation and the combustion emission. The M701F combustion core system and its parameters is considered to be strong coupled nonlinear, therefore a multi-input-multi-output nonlinear modeling system methodology is required to predict or simulate the main parameters in the combustion control system and the gas-fired turbine unit.A BP neural network was applied in the combustion pulsation modeling and the simulation results showed that, the impact law of the pilot ratio on the combustion pressure pulsation at all ranges of frequency varied at different load stage of the gas-fired turbine. When the bypass air valve of the combustion chamber was set, the combustion adjustment goal was reached by fine-tuning the pilot ratio.?...