Study Of Numerical Simulation Based On GPU And Optimization Of The Coal-fired Boilers

As a backdrop for the present work, energy saving and emission reduction has been the most important subject in China. The boiler combustion optimization could improve the boiler efficiency and reduce exhausted pollutant. Howerver, the previous boiler combustion optimization focused on single objective optimization, which could not improve all the objects except the optimized object. To realize the goal of energy saving and emission reduction effectively, the multi-objective optimization research on coal-fired boilers is imperative. Boiler is a complex system, which requires a large amount of data quantity and a new method to simulate. Parallel computing technology based on GPU could get more than 10 times' speedup, which provides a new method for mass data processing and numerical simulation.To keep boilers in safe and stable operation is the primary task. Wether the resistance of powder conveyance pipelines is in balance is very important, for which determines the boiler flame is centralized or not. Presently, a 3d design software called PDMS has been adopted by most electric power design institutes. The software is powerful in the plant design except that it lacks of the function of the resistance computing of powder conveyance pipelines, which should be improved as soon as possible.The contents of this paper are as followings:Firstly, the hardware structure of GPU and the principle of general-purpose computing on graphics processing units are introduced based on the GT200 series GPU. Then, the general-purpose computing on graphics processing units technology developed by Nvidia coporation called compute unified device architecture, which is called cuda for short, is shown.Secondly, the classical problem of flowing aroud a circle cylinder is studied, and the corresponding parallel algorithm based on GPU and LBM is developed. The various factors that affect the performance of the algorithm are discussed in this paper.Thirdly, the support vector machine is used to build the NOx and the carbon burnout models of the 300MW dual-furnace coal-fired boiler. The parallel algorithm of support vector machin based on GPU is introduced, which is effective for mass data processing, however is useless for less data.Fourthly, various multi-objective algorithms called NSGA-â…¡, SPEA2, MOCell, AbYSS and OMOPSO are used for the muti-objective optimization of the boiler. At the same time, the algorithms are compared by the exeute time and the qulity of the optimization results. Fifthly, pipeline resistance computing software is developed base on the PDMS, which could be cooperated with PDMS seamlessly, and could improve the efficiency of the design engineers.Finally, based on the researches above, two arrangment schemes of coal bulk are compared by the resistance characteristics.