Abstract:  In order to solve the problem that the ash Ad, the sulfur fraction Std, the crushing strength M40 and the wear strength M10 in the coke quality index are difficult to measure in real time, an adaptive differential evolution algorithm (ADE) is proposed to optimize the BP network (ADE-BP) coke quality prediction model. Based on the actual input and output index system, the model is trained and simulated according to the historical data of the actual coking production process. The simulation results show that the adaptive differential evolution algorithm to optimize the coke quality model of BP network has higher prediction accuracy. This study provides a new idea for the difficult problem of coke quality index in coking production process, which can provide theoretical basis for high efficiency and low consumption production in coking industry.

Key words: Coke quality, adaptive differential evolution algorithm, BP network, prediction accuracy

摘要： 为了解决炼焦生产过程中焦炭质量指标中的灰分Ad，硫分Std，抗碎强度M40，耐磨强度M10难以实时测量的问题，建立了一种自适应差分进化算法(ADE)优化BP网络(ADE-BP)的焦炭质量预测模型。基于面向实际建立的输入输出指标体系，依据焦化厂实际炼焦生产过程中的历史数据，对模型进行训练和仿真，仿真结果表明，自适应差分进化算法优化BP网络的焦炭质量模型具有更高的预测精度。该研究为炼焦生产过程中焦炭质量指标难以实时监测的难题提供了一种新思路，可为炼焦行业高效低耗生产提供理论依据。

关键词: 焦炭质量, 自适应差分进化算法, BP网络, 预测精度