基于CatBoost-PSO的水泥煤耗优化

Optimization of Cement and Coal Consumption Based on CatBoost-PSO

水泥工业实际生产过程中存在着时变性、时延性,以及生产过程中过多不确定性影响因素的干扰。为了降低水泥实际生产的煤耗,提出了一套数据驱动建模优化工艺参数的方法。首先根据Pearson相关系数和归一化后的方差进行特征选择,接着分别使用随机森林、XGBoost、LightGBM和CatBoost进行吨煤耗预测,然后通过TPE对回归模型参数进行优化,其R^(2)决定系数分别提升了0.009、0.0204、0.0033和0.0017,最后考虑迭代时间确定使用CatBoost结合粒子群算法的方法来预测吨煤耗并优化相关特征参数。最后确定其最低吨煤耗为77.91 kg以及工艺参数最优值。分析了不同参数对煤耗调优的影响程度,其中分解炉用煤率、预热器三次风温、C1出口压力和主传窑转速的影响最大,优化权值均在10%以上。

In the actual production process of the cement industry,there are time-varying and ductile factors,as well as interference from too many uncertain factors in the production process.In order to reduce the actual coal consumption in cement production,a data-driven modeling optimization method for process parameters has been proposed.Firstly,feature selection was performed based on Pearson correlation coefficient and normalized variance.Then,random forest,XGBoost,LightGBM and CatBoost were used to predict coal consumption per ton,and TPE was used to optimize the regression model parameters.The R^(2) determination coefficients were increased by 0.009,0.0204,0.0033 and 0.0017 respectively.Finally,consider the iteration time to determine the use of CatBoost combined with particle swarm optimization algorithm to predict coal consumption per ton and optimize relevant feature parameters.Finally,the minimum coal consumption per ton was determined to be 77.91 kg and the optimal process parameters were determined.The impact of different parameters on coal consumption optimization was analyzed,among which the coal consumption rate of the decomposition furnace,the tertiary air temperature of the preheater,the C1 outlet pressure,and the main kiln speed have the greatest impact,with optimization weights above 10%.