An Application Research Of Optimized Coal Blending Structure Based On Coal Petrology And Immune Genetic Algorithm

The coking coal resources are relatively scarce and uneven distribution in China.Then high-quality coking coal import is increased year by year.At the same time,the coke industry overcapacity is serious and very competitive.Moreover,the quality of the coke and coking cost are put forward higher requirements.However,on the one hand,the coal blending technology based on traditional coal chemical indicators has high dependence on high-quality coking coal,especially high-quality fat coal and coking coal.On the other hand,due to the limitations of traditional coal blending technology,the production targets of high and stable coke quality are difficult to be guaranteed.Meanwhile,the coal blending costs is high.Therefore,the optimization of coal blending structure and optimization method of coal blending based on immune genetic algorithm are studied in this thesis,so as to realize the purpose of saving high quality coking coal,stabilizing coke quality and reducing coal blending costs.At first,the single coals were comprehensively analyzed through coal petrological indicators and coal chemical indicators.Secondly,the vitrinite random reflectance of the blending coal was divided into R1,R2,R3,R4 and R5 indicators,and 5 kg test coke oven stamping coking experiments were carried out,with the five indicators as experimental variables,in order to explore the rationality and feasibility of increasing the coal blending parameters of R1,R2,R3,R4 and R5 in optimizing coal blending structure.Then,the least squares method and the support vector machine(SVM)were used to establish the linear prediction models of sulfur and ash of coke,and the non-linear prediction models of cold strength and thermal strength of coke.At last,based on coal petrological blending technology and coke quality prediction model,the optimization model of coal blending ratio was established by the immune genetic algorithm,and was validated for practicality.Based on the above study,the main experimental conclusions are as follows:(1)There has strong linear correlation between the five indicators and the dryash-free volatile matter,the mean maximum reflectance of vitrinite,the caking index,the maximum thickness of plastic layer,respectively.Moreover,the coke reactivity index improves obviously by the increasing of the content of R1,and reduces by the increasing of the content of R3 and R4.Meanwhile,the degree of impacting on the coke reactivity index by R1,R2,R3,R4 and R5 is R1 > R3 = R4 > R5 > R2.Furthermore,on the condition of meeting the blended coal quality requirements in tamping coking process,1/3 coking coal can instead of fat coal and the proportion of coking coal can be reduced.The thermal strength of coke can be effectively improved by adjusting that the respective content of R1,R2,R3,R4 and R5 are 25% ? 35%,20% ? 30%,10% ? 20%,15% ? 25% and 5% ? 15%.(2)The established coke quality prediction models have high prediction accuracy and strong generalization ability,with using the support vector machine(SVM)method to establish the prediction models of the cold strength and thermal strength of coke,and using the principal component analysis(PCA)method to reduce the dimension of the models inputs,and using the grid search method and K-CV cross validation method to optimize the parameters in the models and verify the generalization ability of the models.(3)Under the conditions of the quality constraints of blending coal and coke,the optimization model of coal blending ratio established by the immune genetic algorithm can stabilize the convergence in the range of 250 ? 300,and obtain the stable and low cost coal blending ratio,and greatly reduce the amount of high-quality fat coal and coking coal.Then,it can achieve the purpose of saving high-quality coking coal,stabilizing coke quality and reducing coal blending costs.