Research On Multi-objective Genetic Algorithm And Its Application To Coal Gasification Process

Our country is rich in coal resources,but the short of oil and gas resources. In recentyears, with the development of IGCC power generation, fuel cell, synthesis gas, syntheticammonia, synthetic methanol, hydrogen production, oil refining and metallurgy, coal andother industrial, the use of coal resources and coal gasification technology have been paidmore attention to. Coal gasification process’s ultimate goal is to generate high quality syngasas much as possible. However since still exists some questions on the optimal control of thecoal gasification operating conditions, coal gasification can lead to serious problems ofresource waste and environmental pollution. Coal gasification process operation parametersoptimization not only can improve the syngas production efficiency, guarantee the quality ofsyngas,but also it can reduce energy consumption and improve resourceu utilization.Therefore,coal gasification process parameters optimization’s research has importanttheoretical and practical significance.In this dissertation,for the fixed bed gasification process,the main results can be summarized as follows:(1) A gasification process prediction model of Shenfu coal fixed bed is established basedon the black-box model of Least Squares Support Vector Machine and BP neural network.The model parameters are solved by genetic algorithm, and furthermore, the functionrelationship between the main evaluation indicators(Cold gas efficiency, Gas yield and Lowheating Value) and the gasification process control parameters (oxygen-coal ratio andcoal-water ratio) is fitted.The results for two kinds of moles for predictiong precision testhave shown that the GA-LSSVM model has better prediction accuracy and generalizationability than the GA-BP model for a gasification process prediction model of Shenfu coal.(2) The multi-objective genetic algorithm (CLS-OB-NSGA-II) is proposed by combiningchaotic local search strategy and algorithm OB-NSGA-II. The individuals on the non-dominated layer1or2of each generation elitism strategy population can be Chaossearched based on individuals in the OB-NSGA-II algorithm. The multi-objective geneticalgorithm NSGA-II, OB-NSGA-II and this improved algorithm (CLS-OB-NSGA-II) havebeen used in test function for two goals and three target separately and the test results showthat: improved multi-objective genetic algorithm has a more satisfactory convergence anddistribution, and is able to converge to the Pareto near the true solution of the optimizationproblems in using a smaller initial population size and less less number of evolution.(3) The multi-objective optimization model of the fixed bed gasification process hasbeen established according to a functional relationship between the key performanceindicators and the main control parameters fitted of the coal gasification process. The Shenfucoal fixed bed gasification process parameters have been optimized by improvedmulti-objective genetic algorithm and the gasification process fitting simulation results showthat the improved multi-objective genetic algorithm (CLS-OB-NSGA-II) which obtains abetter Pareto solution set than the original algorithm OB-NAGA-II, can obtain a Paretosolution set of problems in the smaller population size and less iteration number ofevolution.The optimized parameter values are in the actual fixed bed gasification optimaloperating range, which has proved the feasibility and effectiveness of the improved algorithmapplied to coal gasification process.