基于改进WOA-LSTM的焦炭质量预测

“双碳”背景下,提升焦炭质量是保证钢铁行业高质量发展的研究重点之一,而炼焦行业存在着在线实时监测难、焦炭质量预测模型泛化能力差等问题。为此,提出一种通过自适应全局搜索算法,即改进鲸鱼优化算法（WOA）与长短期记忆（LSTM）循环神经网络综合建模的方法来解决这一问题。首先选取出配合煤中可反映焦炭质量的可测参数,再运用主成分分析（PCA）去除变异性小的冗余因子后,得到预测因子,将其作为LSTM网络的外部输入;通过加入自适应惯性权重以及最佳扰动更新改进WOA,从而训练LSTM网络的超参数,采用均方根误差（RMSE）和R-squared 进行算法检验;最后将改进后的AGWOA-LSTM模型与典型的LSTM、WOA-LSTM模型进行对比,以验证本方法的优越性。结果表明AGWOA-LSTM模型预测焦炭质量具有精度高、运行速度快等特点。研究对焦炭生产具有一定的理论指导意义。

Prediction of coke quality based on improved WOA-LSTM

In the context of “double carbon”, improving coke quality is one of the key points to ensure the high quality development of the steel industry. The coking industry has the problem of on-line real-time monitoring and the generalization ability of the coke quality prediction model is relatively poor. An adaptive global search algorithm, namely improved whale optimization algorithm (WOA) and long short-term memory (LSTM) recurrent neural network integrated modeling method is proposed to solve this problem. We select the measurable parameters that can reflect the coke quality in the blended coal, and use principal component analysis (PCA) to remove the redundancy factors with small variability to obtain the prediction factors as the external input of LSTM network; add adaptive inertia weight and optimal disturbance update to improve WOA, so as to train the super parameters of LSTM network, and use root mean square error (RMSE) and R-squared to test the algorithm. The improved AGWOA-LSTM model is compared with the LSTM model and WOA-LSTM model to verify the superiority of the method. The results show that the AGWOA-LSTM model has high accuracy and fast operation speed, and has a guiding significance for coke production.