改进型决策树在加热炉热效率评估中的应用研究

Research of Applying Improved Decision Tree to Thermal Efficiency Evaluation of Heating Furnaces

在石化企业加热炉设备的长期运转和热效率评估过程中,针对属性数量众多且多为连续型属性这一特点,利用传统规则库的调炉决策很难对加热炉进行有效的能效评估。因此,提出了一种基于C4.5算法的改进型分类算法,弥补C4.5算法没有考虑问题无关属性和冗余属性对分类结果的影响。在经典的C4.5算法基础上引入模拟退火算法对分类属性进行优化组合,使得分类算法在分类时间和准确率上都有改进和提升。同时,在已有的炉况在线监测系统上加以改进,将决策树算法运用到该系统中,用于石化加热炉热效率的评估,为管理人员提供炉况调整解决方案。

In the process of long-term operation and thermal efficiency evaluation of heating furnaces in petrochemical enterprises, the traditional rule base has difficulty in evaluating the energy efficiency of heating furnaces effectively. Therefore, an improved classification algorithm based on C4.5 algorithm was proposed to compensate the C4.5 algorithm which does not consider the impact of problem-independent attributes and redundant attributes on the classification results. Based on the classical C4.5 algorithm, the simulated annealing algorithm was introduced to optimize the combination of classification attributes so that the classification algorithm can be improved in both classification time and accuracy. At the same time, the existing on-line monitoring system for furnace condition was improved and the decision tree algorithm was applied to evaluating the heating efficiency of the petrochemical heating furnace and to provide a solution for managers to adjust furnace conditions.