【摘要】 为提高采煤工作面涌水量预测准确度，收集大量工作面涌水量观测数据进行整理、统计、分析，将涌水量稳定性、周期性和季节性特征考虑在内，提出1种基于数据驱动的完全自适应模态分解算法(CEEMDAN)和改进的混合时间序列模型工作面涌水量预测方法。该方法利用CEEMDAN处理涌水量数据，构建麻雀搜索算法(SSA)优化的长短期记忆网络(LSTM)和自回归移动平均模型(ARIMA)并行级联而成的混合时间序列模型对工作面涌水量进行预测。研究结果表明：该模型预测结果与真实数据相差更小，平均绝对误差为6.36 m~3/h,均方根误差为10.6 m~3/h,模型拟合系数为0.95,更适用于工作面涌水量预测。研究结果可为矿井工作面涌水量预测及防控提供参考。更多还原

【Abstract】 In order to improve the prediction accuracy of water inflow in coal mining face, a large number of observation data of water inflow in coal mining face were collected for collation, statistics and analysis.Taking into account the stability, periodicity and seasonal characteristics of water inflow, a prediction method of water inflow in working face based on the data-driven adaptive noise-complete set empirical mode decomposition algorithm(CEEMDAN) and the hybrid time series model was proposed.In this method, the water inflow data was processed by using CEEMDAN,and a hybrid time series model formed by the parallel concatenation of long short-term memory network(LSTM) optimized by sparrow search algorithm(SSA) and autoregressive integrated moving average model(ARIMA) was constructed to predict the water inflow of working face.The results show that the difference between the prediction results of the hybrid model and the real data is smaller, and it is more suitable for the prediction of water inflow in working face.The average absolute error of the model prediction results is reduced to 6.36 m~3/h, the root mean square error is reduced to 10.6 m~3/h, and the model fit coefficients are 0.95,which not only overcomes the interference of other related influencing factors, but also improves the prediction accuracy and speeds up the prediction speed.The research results can provide a reference for the prediction and prevention of water inflow in mine working faces.更多还原