面向吨钢综合能耗预测的基因表达式编程方法

Using Gene Expression Programming to Predict Integrate Energy Consumption per Ton Crude Steel

钢铁是国民经济的基础用材,我国吨钢综合能耗仍与发达国家存在较大差距。准确预测吨钢综合能耗,有利于制定节能方针和减少能源浪费。根据我国钢铁工业吨钢综合能耗历史数据,利用基因表达式编程(Gene Expression Programming,GEP)算法,构建吨钢综合能耗预测模型。首先,将吨钢综合能耗进行等间隔时序化、函数表达式符号化,在终端集中添加常量数组;其次,利用选择操作、变异操作、重组操作和移项操作进行遗传操作,获得吨钢综合能耗预测模型。结果表明,基于GEP的预测值与实测值平均误差为0.31,该模型较准确地预测我国钢铁工业吨钢综合能耗发展趋势。

The iron and steel industry is a fundamental part of national economy, but it is also the large energy user. If possible, the predictable accuracy is helpful to reduce the energy waste and predict the energy development trends. An improved Gene Expression Programming （GEP）/s proposed to construct the prediction model of integrate energy consumption per ton crude steel according to the our country＇s historical data. Firstly, the integrate energy consumption per ton crude steel is divided into equal time intervals. Some symbols are defined to represent the funetion. In the terminal set, a constant array is added. Secondly, the prediction model is obtained by genetic operation, which contains selection operation, mutation operation, recombination operation, transposition operation, etc. Finally, the experimental results show that the mean deviation is 0.31 via GEP, which is the minimum one.