基于改进果蝇算法和长短期记忆神经网络的油田产量预测模型  

Oilfield Production Prediction Model Based on Improved Fruit Fly Algorithm and Long-short Term Memory Neural Network

作　　者：任燕龙 谷建伟[1] 崔文富[2] 张以根[3] REN Yan-long;GU Jian-wei;CUI Wen-fu;ZHANG Yi-gen(School of Petroleum Engineering,China University of Petroleum(East China),Qingdao 266580,China;Shengli Oil Production Plant,Shengli Oilfield Company,SINOPEC,Dongying 257015,China;Exploration and Development Research Institute,Shengli Oilfield Company,SINOPEC,Dongying 257015,China)

机构地区：[1]中国石油大学(华东)石油工程学院,青岛266580 [2]中国石化胜利油田分公司勘探开发研究院,东营257015 [3]中国石化胜利油田分公司胜利采油厂,东营257015

出　　处：《科学技术与工程》2020年第18期7245-7251,共7页Science Technology and Engineering

基　　金：国家科技重大专项(2017ZX05009001)。

摘　　要：产量预测是油田生产动态开发研究的重要内容之一。油田的长期生产积累了大量数据,但是波动幅度很大,直接应用长短期记忆神经网络预测油田的生产指标,会出现神经网络泛化性很差的问题。因此,首先利用双层长短期记忆神经网络(long-short term memory,LSTM)和随机式失活对神经网络架构进行调整,建立了深度学习神经网络模型;并提出了一种新的果蝇聚集方法,通过改进的果蝇优化算法对所建立的神经网络模型进行优化,避免其陷入局部最优解,搜寻解空间的最优解;最后,油田实例验证表明,优化后的深度学习网络的网络泛化能力和预测精度有了较大提高,对于油田波动性较大的数据也能较好地拟合。所建立油田产量预测模型可应用于矿场开发实际。Production prediction is one of the important tasks of dynamic research and development of oilfield production. Although the long-term production of oil fields can accumulate a large amount of data, the fluctuation range is very large. Thus, if the long-short term memory neural network are directly applied to predict the production indicators of oil fields, the generalization of neural networks would be very poor. Therefore, the two-layer long-short term memory(LSTM) and random inactivation was used to adjust the neural network architecture. A deep-learning neural network model was established and optimized by fruit fly optimization algorithm, an improved fruit fly aggregation method. The new method could avoid from falling into the local optimal solution and search for the optimal solution in the solution space. Finally, the field example verification showed that the network generalization ability and prediction accuracy of the optimized deep-learning network were greatly improved, and the data with large fluctuations in the oilfield were well fitted. Therefore, the proposed oilfield production prediction model can be applied for the actual development in the field.

关 键 词：果蝇算法 浓度聚集 长短期记忆网络 随机失活 深度学习 产量预测 

分 类 号：TE331[石油与天然气工程—油气田开发工程]