Method for performing nonlinear prediction on coke quality on basis of cohesiveness and coal-rock indexes of single coal

Abstract

The invention discloses a method for performing nonlinear prediction on coke quality on the basis of cohesiveness and coal-rock indexes of single coal, which provides important technical guarantee for improving the stability and quality of coke produced by coke making enterprises. The method comprises the following steps of: establishing an information database storing qualities and coke quality indexes of the single coal for coking, and inputting the cohesiveness indexes and the coal-rock indexes of the single coal for coking into the coal information database; establishing a coal quality prediction model, and predicting quality indexes of matched coal by virtue of a clustering and support vector machine; defining the quality indexes of the matched coal, and predicting quality indexes of the coke according to the quality indexes of the single coal; and establishing a model for predicting the crushing strength and abrasive resistance of the coke. The invention further provides a coke quality prediction system with optimized coal blending of coal-rock, which aims to stabilize and improve the coke quality and reduce the coal blending cost, can form a prediction model with multiple parameters and high accuracy for the coal quality indexes and the coke quality indexes, and simultaneously, has a real-time updating function or a manual intervention function.

Description

technical field [0001] The invention belongs to the coal chemical industry technical field coking industry coking production process coke quality prediction method, specifically relates to coking single coal quality cohesion index level, including two factors of the maximum thickness Y value of colloidal layer and cohesion index G value; coal Rock index level, including two factors of vitrinite total reflectance and microcomponent activity-inert ratio, constitutes a practical system of two levels and four factors to predict coke mechanical strength and thermal performance. In particular, it is a method for predicting coke quality nonlinearly from a single coal caking and coal-rock index. Background technique [0002] The large-scale blast furnace and the improvement of oxygen-coal injection technology put forward higher requirements on the quality and stability of coke used in blast furnaces. In the current situation of various types of coking coal and unstable coal quality...

AI Technical Summary

Problems solved by technology

Chinese patent CN1749358A discloses the use of multiple linear regression methods to predict coke quality indicators. This method has the characteristics of simple methods, but when this method is predicted, it is difficult to obtain deterministic values ​​because of the linear regression relationship; The quality is complex, and the volatile content of coking coal under the same conditions V daf and G value, there will be a big difference in predicting the mechanical strength and thermal properties of coke, because after a single coal is mixed, different quantities and types of coal can be used to obtain the same volatile V daf and G value

At the same time, because coking coal and coke quality indicators are a complex relationship in the coking process, using a linear relationship to replace the nonlinear relationship in the coking process of coking coal will inevitably increase the error between the predicted results of the coke quality indicators and the actual ones.

Chinese patent CN101661026A uses BP neural network to predict coke quality. In fact, this method uses the principle of empirical risk minimization, but the actual result cannot minimize the expected risk. There are theoretical flaws, and it is easy to fall into local minimum points, and the generalization ability Not strong and other shortcomings

In addition, the number of hidden layers and the number of nodes in the hidden layer of the neural network are generally determined by experience, and there is no clear algorithm, which affects the prediction accuracy of the neural network to a certain extent, and is subject to certain restrictions in practical applications.

Images