Coal blending expert system fit for top-loading coke oven

Abstract

The invention discloses a coal blending expert system fit for a top-loading coke oven. The coal blending expert system comprises a coal data and coal blending data acquisition module, a coke performance autoregression module, an optimization coal blending module and a coal blending result acquisition and inspection module. The system adopts a combined method of multi-parameter nonlinear regression, a coal vitrinite reflectance distribution diagram and a coal fluidity distribution diagram, so that the coke strength prediction deviation is controlled to be within 2%. Meanwhile, nonlinear optimization coal blending is carried out with the goal of the lowest coke net cost, the influence of the coking byproduct yield, the market price and the coal quality and price on the net cost is considered, and coking and coal blending with the lowest cost are really carried out.

Description

technical field [0001] The application belongs to the field of coking coal blending, in particular to a coal blending expert system suitable for top-loaded coke ovens. Background technique [0002] With the decline of coke profits, various coal blending coking systems have emerged for the purpose of reducing costs. When the time, coking temperature and coke quenching method change, the deviation of the prediction results is relatively large. Patent ZL201010576330.7 adopts regression method to fit Romax, fluidity MF, ash content and other indicators to predict coke quality, but this model only considers coal quality, but does not consider coke oven coking time, coking temperature, and coke quenching The impact of the method on the quality of coke; the patent 200810013054.6 BP neural network method to predict the quality of coke, the vitrinite reflectance distribution and cohesiveness index of coal as the input section of the parameter input network, and the corresponding cok...

AI Technical Summary

Problems solved by technology

[0002] With the decline of coke profits, various coal blending coking systems have emerged for the purpose of reducing costs. When the time, coking temperature, and coke quenching method change, the prediction results will deviate greatly

Patent ZL201010576330.7 adopts regression method to fit Romax, fluidity MF, ash content and other indicators to predict coke quality, but this model only considers coal quality, but does not consider coke oven coking time, coking temperature, and coke quenching The impact of the method on the quality of coke; the patent 200810013054.6 BP neural network method to predict the quality of coke, the vitrinite reflectance distribution and cohesiveness index of coal as the input section of the parameter input network, and the corresponding coke index as the BP neural network At the output end, the BP neural network is trained to obtain the nonlinear relationship between input parameters and output parameters, and then the index to be predicted is used as the input parameter of the BP neural network to obtain the predicted strength of coke. This method ignores the ash and volatile content of coke , maximum colloidal layer thickness, coke oven operating parameters and other indicators lead to deviations in coke prediction, and the neural network method is more accurate in predicting coke quality for used coal types, but the prediction deviation for new coal types is relatively large

Images