基于SG-LSTM-GRU井下粉尘体积分数预测模型
Prediction of underground dust concentration based on SG-LSTM-GRU model
牛莉莉 1杨超宇1
作者信息
- 1. 安徽理工大学 经济与管理学院,安徽 淮南 232001
- 折叠
摘要
粉尘体积分数监测对煤矿粉尘预警起着重要的作用,对其变化预测有利于保障井下安全生产和降低井下矿工职业病风险.针对井下粉尘体积分数预测问题,建立了一种SG-LSTM-GRU预测模型.对监测数据进行预处理,先采用SG滤波方法处理,可以减少时间序列数据中存在的噪声数据,再用最大最小值法进行归一化处理,将得到的数据集划分为90%训练集与10%测试集,通过枚举法不断地实验确定SG滤波器中的参数值以及模型中时间步长,再通过 LSTM 和 GRU 模型相结合的方式,既可以通过LSTM层保留数据中较多的特征,又可以通过GRU层提高模型的训练速度,采用均方根误差、平均绝对误差和计算时间进行模型对比评价.研究结果表明,SG-LSTM-GRU模型预测的RMSE为0.256,MAE为 0.066,相比其他预测模型该模型预测效果好.因此,采取SG-LSTM-GRU模型对井下粉尘体积分数预测,可以提高数据预测准确性,实现煤矿井下安全生产以及降低矿工尘肺病风险.
Abstract
Dust concentration monitoring plays an important role in coal mine dust warning,and predicting its concentration changes is beneficial for ensuring underground safety production and reducing occupational disease risks for miners.A SG-LSTM-GRU prediction model had been established for the prediction of underground dust concentration.Preprocess the monitoring data by first using the SG filtering method to reduce the presence of noise in the time series data.Used the maximum and minimum value method for normalization.Divide the obtained dataset into a 90%training set and a 10%testing set.Continuously experiment through enumeration to determine the parameter values in the SG filter and the time step size in the model.Combined LSTM and GRU models.Both the LSTM layer could retain more features in the data,and the GRU layer could improve the training speed of the model.Root mean square error,mean absolute error,and calculation time were used for model comparison and evaluation.The research results indicated that the RMSE and MAE predicted by the SG-LSTM-GRU model were 0.256 and 0.066,which was better than other prediction models.Therefore,adopting the SG-LSTM-GRU model for predicting underground dust concentration could improve the accuracy of dust concentration data prediction,achieve safe production in coal mines underground,and reduce the risk of pneumoconiosis among miners.
关键词
机器学习/LSTM/GRU/Savitzky/Golay滤波器/粉尘体积分数预测/安全生产Key words
machine learning/LSTM/GRU/Savitzky Golay filter/dust concentration prediction/safe production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