摘要
针对现阶段油田产量预测中所出现的一些预测效果不理想的问题,开展了对全国原油产量的时序预测研究.针对全国原油2011-2020年产量所呈现出来的特点,采用一种基于时间序列自回归移动平均模型(ARIMA)结合长短期记忆网络(LSTM)组合模型的预测方法.首先,运用时间序列ARIMA模型的建模思想,对全国原油产量进行初步预测,再通过LSTM训练拟合残差并进行预测.最后将LSTM的预测结果补偿到初步预测结果中,得到组合预测值.组合模型预测结果显示,预测结果比较可靠,对预估原油产量具有一定的参考价值.
Aiming at some unsatisfactory forecasting problems in the current oilfield production forecasting,this article has carried out a time series forecasting study of national crude oil production.In this project,aiming at the characteristics of national crude oil production from 2011 to 2020,a forecasting method based on time series autoregressive moving average model(ARIMA) combined with long short-term memory network(LSTM) combined model is adopted.First,use the modeling ideas of the time series ARIMA model to make a preliminary prediction of the national crude oil production,and then use LSTM training to fit the residuals and make predictions.Finally,the prediction result of LSTM is compensated to the preliminary prediction result,and the combined prediction value is obtained.The prediction results of the combined model show that the prediction results are relatively reliable and have a certain reference value for the estimated crude oil production.
作者
李阳
杜睿山
程永昌
LI Yang;DU Rui-shan;CHENG Yong-chang(School of Mathematics and Statistics,Northeast Petroleum University,Daqing 163318,China;School of Computer and Information Technology,Northeast Petroleum University,Daqing 163318,China)
出处
《数学的实践与认识》
2022年第6期40-48,共9页
Mathematics in Practice and Theory
基金
东北石油大学引导性创新基金(2020YDL-04)。
