基于多重多元回归的焦炭质量预测模型

 焦炭质量预测是焦化企业进行焦炭质量控制的重要方法。在诸多生产因素已固定的条件下，焦炭质量主要取决于原料煤性质。用煤质指标预测焦炭质量是焦炭质量预测的重要方法。考虑配合煤煤质指标和焦炭质量指标的多元性和相关性，采用多重多元回归分析技术，将焦炭质量的各指标作为一个整体，建立配合煤煤质指标对焦炭质量的预测模型。基于偏最小二乘回归思想，采用预测误差平方和（PRESS）和预测的方差验证回归模型的预测能力。根据实际焦炭生产建立的焦炭质量预测多重多元回归模型的显著性检验表明，该模型具有较高的预测能力。应用多重多元回归技术建立的焦炭质量预测模型对指导焦化企业的焦化生产，优化配煤和加强焦炭质量控制具有重要的现实意义。

Coke Quality Prediction Model Based on Multivariate Regression

The prediction of coke quality in a coke quality controlling process is important for coke production facilities. With many production conditions such as equipments, parameters, fixed, the coke quality mainly depends on the quality of coal. Using the quality index of coal to forecast the quality of coke is an important method of coke quality prediction. Because the relations between quality indexes of the blended coal and the coke take a multiple nature and are mutually related, so all quality indexes of the coke are taken as a whole and then the multivariate regression is used to establish the coke quality prediction model. In order to assure the model a good performance in predicting, firstly, the fitting accuracy of the regression model is tested, then, the Prediction Error Square Summation (PRESS) and Prediction Variance are used to test the model based on the partial least-squares algorithm. According to a practical coke production, a coke quality prediction model based on multivariate regression is built. The significant test shows that the model is correct and the predicting test shows that this model is accurate in prediction. The practical application of the method shows that the coke quality prediction model based on multivariate regression is reliable for coke production in blending coal optimally and in coke quality control.