摘要
为解决煤矿瓦斯浓度预测问题,提出基于多因素的LSTM瓦斯浓度预测模型。模型首先对煤矿多源监测数据进行数据融合、缺失值处理;其次通过特征衍生、有监督化、无量纲化,融合各环境因素特征和时序数据的时间性特征,且衍生出更多交叉项特征和高次项特征;然后利用经验法和逐步试错法确定隐藏层维度;最后进行模型训练和测试。研究结果表明:基于多因素的LSTM瓦斯浓度预测模型的RMSE仅为0.021,MAE为0.01,比单因素LSTM模型、RNN模型预测效果好。因此,基于多因素的LSTM瓦斯浓度预测模型可更精准地进行瓦斯浓度多步预测,促进煤矿安全生产。
In order to solve the problem of gas concentration prediction in coal mine,a LSTM gas concentration prediction model based on multiple factors was put forward.Firstly,the data fusion and missing value processing of multi-source monitoring data in coal mine were conducted.Secondly,through the feature derivation,supervision and dimensionless,the characteristics of various environmental factors and the temporal characteristics of time series data were fused,and more cross-term features and high-order features were derived.Thirdly,the empirical method and stepwise trial-and-error method were used to determine the hidden layer dimension.Finally,the model was trained and tested.The results showed that the RMSE of LSTM gas concentration prediction model based on multiple factors was only 0.021,and MAE was 0.01,so the prediction performance was better than that of single-factor LSTM model and RNN model.Therefore,the LSTM gas concentration prediction model based on multiple factors can more accurately predict the gas concentration in multiple steps and promote the work safety of coal mines.
作者
刘莹
杨超宇
LIU Ying;YANG Chaoyu(School of Economics and Management,Anhui University of Science and Technology,Huainan Anhui 232001,China)
出处
《中国安全生产科学技术》
CAS
CSCD
北大核心
2022年第1期108-113,共6页
Journal of Safety Science and Technology
基金
国家自然科学基金项目(61873004)。
关键词
LSTM
瓦斯浓度预测
数据融合
时间序列
特征衍生
long short-term memory(LSTM)
gas concentration prediction
data fusion
time series
feature derivation
作者简介
刘莹,硕士研究生,主要研究方向为自然语言处理、数据挖掘。
