基于麻雀优化算法和CNN-BiLISTM的矿压预测模型
Mine Pressure Prediction Model Based on Sparrow Optimization Algorithm and CNN-BiLISTM
束云龙 1张华磊1
作者信息
- 1. 安徽理工大学矿业工程学院,安徽 淮南 232000
- 折叠
摘要
为了精准预测工作面顶板来压步距,提出了一种智能预测模型,将融合正余弦和柯西变异的麻雀优化算法(SCSSA)与卷积双向长短期记忆神经网络(CNN-BiLSTM)相结合,从而实现工作面顶板来压步距的精确预测.实验结果表明,SCSSA-CNN-BiLSTM来压步距预测模型预测值与实测值基本吻合,RMSE为1.051,MSE为 1.104 5,MAE为 0.816 2,MAPE为 0.056 6,相关系数R为 0.994 3,决定系数R2 为 0.949 7,可以完成对于来压步距的预测.以样本1 举例,实际值为 20.20,而CNN-BiLSTM模型预测值为 15.22,SCSSA-CNN-BILSTM模型预测值为 18.49;以样本 10 举列,实际值为 9.55,而CNN-BiLSTM模型预测值为 13.78,SCSSA-CNN-BILSTM模型预测值为 10.60,与未经过优化的CNN-BiLSTM算法相比,预测精度提升了约18%,参数优化效果显著提升.
Abstract
In order to accurately predict the weighting step distance of the working face roof,an intelligent prediction model is proposed,which combines the Sparrow Optimization Algorithm(SCSSA)that integrates sine and cosine and Cauchy variation with the Convolutional Bidirectional Long Short Term Memory Neural Network(CNN-BiLSTM)to achieve accurate prediction of the weighting step distance of the working face roof.The experimental results show that the predicted values of the SCSSA-CNN-BiLSTM weighting step distance prediction model are basically consistent with the measured values,with RMSE of 1.051,MSE of 1.104 5,MAE of 0.816 2,MAPE of 0.056 6,correlation coefficient R of 0.994 3,and determination coefficient R2 of 0.949 7,which can complete the prediction of weighting step distance.Taking Sample 1 as an example,the actual value is 20.20,while the predicted value of the CNN-BiLSTM model is 15.22,and the predicted value of the SCSSA-CNN-BILSTM model is 18.49;taking sample 10 as an example,the actual value is 9.55,while the predicted value of the CNN-BiLSTM model is 13.78,and the predicted value of the SCSSA-CNN-BILSTM model is 10.60.Compared with the unoptimized CNN-BiLSTM algorithm,the prediction accuracy improves by about 18%,and the parameter optimization effect significantly improves.
关键词
矿压/步距/预测/模型Key words
mine pressure/step distance/prediction/model引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据