Research On Fast Learning Network And Genetic Algorithm And Applications In Heat Transfer Optimization Of Cement Clinker

Fast Learning Networks(FLN)is a double parallel non-iterative learning neural network with fast learning speed and good generalization performance;Non-dominated Sorting Genetic Algorithm II(NSGA-II)is a multi-objective intelligent optimization algorithm developed by genetic algorithm.They can solve the modeling and optimization problems of complex systems and have been applied in many fields.In this paper,FLN and NSGA-II have been deeply researched and improved,and they are applied to the model identification and parameter optimization of the grate cooler.Then,combined with the principle of clinker heat transfer in the grate cooler,the heat transfer parameters of cement clinker are optimized.It not only has important theoretical guiding significance for the improvement of heat exchange efficiency of the grate cooler,but also has broad application prospects and great social benefits.The main research contents of this paper are described as follows:Firstly,in order to improve the nonlinear mapping ability of the original FLN,the kernel method of FLN is derived.For the problem of high dimensionality in the large sample set,the reduced kernel matrix is used instead of the derived kernel matrix.A reduced kernel fast learning network(RKFLN)is proposed,and a non-iterative training method based on Lagrangian multiplier method is derived.Then,in order to improve the repeatability and generalization ability of RKFLN training,a non-iterative training algorithm is derived based on sparse Bayesian regression,and a sparse Bayesian reduced kernel fast learning network(SBRKFLN).Simulation test verified the effectiveness and superiority of SBRKFLN.Recursive least squares based echo state networks(RLS-ESN)is proposed based on recursive least squares method,and its effectiveness is verified by numerical simulation of nonlinear autoregressive moving average(NARMA)model.Secondly,aiming at the problem of population precocity and uneven distribution of individuals caused by insufficient diversity of NSGA-II,an adaptive multi-population nondominated sorting genetic algorithm II(AMP-NSGA-II)is proposed.A single population is divided into multiple sub-populations.Coordination and competition processes between multiple groups are designed to maintain individual diversity.Allow dominant populations to gain more opportunities to multiply to simulate population competition in the real world.There is an attachment relationship between disadvantaged individuals and dominant individuals to mimic the symbiosis of species in the real world.The designed evolution of the population enables the individual to be more evenly distributed while maintaining individual diversity.Multi-objective optimization benchmark problems are used to verify that the AMP-NSGA-II has better convergence performance and strong adaptability.Thirdly,in order to optimize the parameters of the grate cooler,the working principle of the grate cooler and the influence of various parameters were qualitatively analyzed.Based on the second law of thermodynamics,the irreversible loss during the heat exchange process of clinker in grate cooler was analyzed.Based on the preset parameters,the proposed AMP-NSGA-II is used to optimize the entropy production of the grate cooler,and the Pareto frontier of energy loss is obtained.Finally,based on the proposed SBRKFLN and RLS-ESN,the identification model of secondary air temperature and exported clinker temperature are established.Experiments show that the proposed algorithm can accurately predict the secondary air temperature and clinker temperature.Based on the established identification models,AMP-NSGA-II is used to optimize the parameters of the grate cooler,and the Pareto frontier of energy loss is used to determine the optimal parameters of the grate cooler.The grate cooler operates under the optimal heat exchange parameters,thereby achieving the purpose of improving the thermal efficiency of the grate cooler and improving energy utilization.