首页| Well performance prediction based on Long Short-Term Memory (LSTM) neural network
Well performance prediction based on Long Short-Term Memory (LSTM) neural network
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NSTL
Fast and accurate prediction of well performance continues to play an increasingly important role in development adjustment and optimization. It is now possible to predict performance more accurately using neural networks thanks to the advancement of artificial intelligence. In this study, A Long Short-Term Memory (LSTM) neural network model which considered gas injection effect was established to forecast the production performance of a carbonate reservoir in the Middle East. Over 12 years of surveillance data from 17 producers and 11 injectors were selected as the dataset. A correlation analysis was performed to determine the input and output variables of the model before establishing the model. Using historical data from the first 4000 days, the model is trained and validated before it is used to predict the performance of the next 500 days. After that, the calculation results of this model and traditional reservoir numerical simulation (RNS) were compared under the same conditions. The results show that the average error of the LSTM method is 43.75% lower than that of traditional RNS. Moreover, the total CPU time and comprehensive computing power consumption of LSTM method only account for 10.43% and 36.46% of RNS's, respectively. Thus, it is clear that the LSTM approach has a significant advantage when it comes to calculating. In the end, we categorized all 17 producers into three groups based on GOR predictions for the next 500 days, and proposed optimization and adjustment techniques for each type. This study provides a new direction for the application of artificial intelligence in oil and gas development.
Performance predictionLong short-term memoryNeural networkTime series dataCarbonate reservoir
Ruijie Huang、Chenji Wei、Baohua Wang
展开 >
Research Institute of Petroleum Exploration and Development, PetroChina, Beijing, 100083, China
NSTL
