程磊, 李正健, 贺智勇,等.基于MIV-PSO-BPNN的掘进面风温预测方法[J].河南理工大学学报（自然科学版）,2024,43(6):11-17.

CHENG L, LI Z J, HE Z Y,et al.Prediction method of heading face wind temperature based on MIV-PSO-BPNN[J].Journal of Henan Polytechnic University(Natural Science) ,2024,43(6):11-17.

基于MIV-PSO-BPNN的掘进面风温预测方法

程磊1,2, 李正健2, 贺智勇3, 史浩镕1, 王鑫1

1.河南理工大学 安全科学与工程学院，河南 焦作  454000；2.煤炭安全生产与清洁高效利用省部共建协同创新中心，河南 焦作  454000；3.河南省科学院，河南 郑州  450046

摘要: 目的 为防治矿井热害，解决矿井掘进面风温预测问题，  方法 提出一种MIV算法优化的PSO-BPNN预测模型。通过利用MIV算法确定模型的输入变量，以BP网络建模，使用粒子群优化算法结合BP神经网络实现掘进工作面风流温度的预测，得到预测结果并与BPNN模型、PSO-BPNN模型、SVR模型相比较。  结果 结果表明：MIV-PSO-BPNN预测模型的相对误差为0.47%~1.81%，分别优于PSO-BPNN、BPNN、SVR预测模型的-3.96%~1.93%，-5.54%~2.98%，-2.16%~2.95%，预测模型的误差为 -0.1~0.5 ℃，表明预测值与实测值基本一致；与BPNN预测模型、PSO-BPNN预测模型、SVR预测模型相比，MIV-PSO-BPNN预测模型的预测结果平均绝对误差分别减少65%，54%，50%，均方误差分别减少88%，78%，69%，表明该预测模型的预测效果优于其他3种模型。  结论 所提模型适用于矿井掘进工作面风温的预测。

关键词:BP神经网络;MIV算法;粒子群优化算法;风温预测;算法优化

doi:10.16186/j.cnki.1673-9787.2023070025

基金项目:国家自然科学基金资助项目（U1904210）

收稿日期:2023/07/15

修回日期:2023/09/08

出版日期:2024-09-24

Prediction method of heading face wind temperature based on MIV-PSO-BPNN

CHENG Lei1,2, LI Zhengjian2, HE Zhiyong3, SHI Haorong1, WANG Xin1

1.College of Safety Science and Engineering，Henan Polytechnic University，Jiaozuo  454000，Henan，China；2.Collaborative Innovation Center of Coal Work Safety and Clean High Efficiency Utilization，Jiaozuo  454000，Henan，China；3.Henan Academy of Sciences，Zhengzhou  450046，Henan，China

Abstract: Objectives To prevent and control thermal damage in mines and solve the problem of predicting wind temperature in mining excavation faces,  Methods a PSO-BPNN prediction model optimized by MIV algorithm was proposed. The MIV algorithm was used to determine the input variables for the model, followed by BP network modeling. The particle swarm optimization algorithm combined with the BP neural network was then employed to predict the airflow temperature of the excavation working face. The predicted results were compared with those from the BPNN model, PSO-BPNN model, and SVR model. Results The results showed that the relative error range of the MIV-PSO-BPNN prediction model was-0.47% to 1.81%, which was superior to the PSO-BPNN, BPNN, and SVR prediction models with ranges of -3.96% to 1.93%, -5.54% to 2.98%, and -2.16% to 2.95%, respectively. The prediction error was between -0.1 ℃ and 0.5 ℃, indicating that the predicted values and tested values were basically consistent; Compared to the BPNN, PSO-BPNN, and SVR prediction models, the MIV-PSO-BPNN model’s average absolute error decreased by 65%, 54%, and 50%, respectively， and the mean square error had decreased by 88%, 78%, and 69%, respectively. This demonstrated  that the prediction effect of the MIV-PSO-BPNN model was superior to the other three models.  Conclusions The proposed model was suitable for predicting the air temperature of mining excavation working faces.

Key words:BP neural network;MIV algorithm;particle swarm optimization algorithm;airflow temperature prediction;algorithm optimization