基于组合模型的产量动态预测方法研究

Dynamic prediction of oil production based on combined model

产量递减预测模型是油气藏工程研究中的一种重要手段。鉴于目前常用的Aprs、GM(1,1)等模型预测条件的局限性,提出一种新的组合模型。首先对原始产量数据进行平滑处理后,将其分解成趋势性跟随机性两部分,用更加适合于递减数据预测的GOM(1,1)模型进行趋势性部分数据预测,而后引入RBF网络对GOM(1,1)模型残差进行修正,最后得到组合模型预测结果。经实例验证,对于呈递减趋势的产量数据,该组合模型的精度与传统单一方法相比有明显提高,适合于工程研究应用。

Production decline prediction model is important in reservoir engineering.In view of the limitation of currently used Aprs and GM(1,1) model,a new combined model is presented.The new model firstly smoothes the original production data by decomposing them into trend terms and randomness.GOM(1,1) model,which is more suitable for forecasting decline data,can be used for trend terms,then RBF network is introduced to amend the residual error of GOM(1,1) model.Finally,we obtain the prediction result of the combined model.It is verified by practical cases that,for production data in decline trend,the accuracy of the combined model is superior to traditional unitary method,and this model is fit for engineering study.