Abnormal pattern recognition and early warning of water flooding in fractured-vuggy reservoir based on LSTM
-
摘要: 大缝大洞的存在和频繁的工作制度导致缝洞型油藏含水率变化特征多样，暴性水淹预警难度大。针对传统预警方法存在的时滞性问题，采用K线理论刻画含水率生产指标变化趋势，总结出充沛型、突破型、反转型等水淹前异常模式；由于循环神经网络能够刻画生产数据间的长程相关性，采用基于循环神经网络的长短期记忆网络（LSTM）自动识别水淹异常模式特征实现暴性水淹预警。仿真实验表明基于LSTM的水淹异常模式识别模型通过变换数据尺度，较好地捕获暴性水淹前数据的整体变化趋势，识别精度明显优于支持向量机、朴素贝叶斯等模型。K线理论刻画的各类异常模式有效解决了传统预测的时滞难题，提前1~3周实现水淹预警，可以为缝洞型油藏水淹预警研究提供新的研究思路。Abstract: The existence of large fractures and caves and frequent working system adjustments have resulted in the diverse characteristics of water cut in the fractured-vuggy reservoir, which makes it difficult to early warn the water flooding.Aiming at the problem of time delay of traditional early warning methods, this paper uses K-line theory to describe the change trend of water cut production indicators, and summarizes the pre-flooding abnormal patterns such as abundant type, breakthrough type and reversal type.Since the recurrent neural network can memorize the long-term correlation between production data, LSTM is used to automatically identify the features of abnormal pattern to realize early warning of water flooding.The experimental results show that by transforming the data scale, the proposed abnormal pattern recognition model based on LSTM can successfully extract the overall trend of data before water flooding.The recognition accuracy of the proposed model is significantly higher than that of support vector machine and naïve Bayes and other models.Various kinds of abnormal patterns described by K-line can effectively solve the traditional problem of prediction delay.The proposed model realizes early warning of water flooding one to three weeks in advance, and provides new ideas for early warning of water flooding.
-
Key words:
- fractured-vuggy reservoir /
- early warning of water flooding /
- K-line theory /
- abnormal pattern /
- LSTM
-
图 1 缝洞型油藏底水锥进过程示意图[6]
Figure 1. Schematic diagram of bottom water coning process in fractured-vuggy reservoir
图 4 基于LSTM的缝洞型油藏水淹异常模式识别及预警框架
Figure 4. Abnormal pattern recognition and early warning framework of water flooding in fracture-cavity reservoir based on LSTM
图 5 基于K线理论和滑动窗口的数据采样示意图
Figure 5. Diagram of data sampling based on K-line theory and sliding window
图 7 各个模型水淹异常模式识别准确率
Figure 7. Recognition accuracy of water flooded abnormal patterns of each model
图 8 S80井2007年10月4日前后含水率生产曲线图(a)和K线变化图(b)
Figure 8. Water cut curve (a) and K-line change (b) of Well S80 around October 4, 2007
图 9 S80井2010年11月13日前后含水率生产曲线(a)和K线变化图(b)
Figure 9. Water cut curve (a) and K-line change(b) of well S80 around November 13, 2010
表 1 缝洞型油藏水淹异常模式
Table 1. Abnormal pattern of water flooding in fractured-vuggy reservoir
水淹异常模式 K线图形态 物理意义 充沛型 
单位时间内含水率呈快速、持续上升趋势，表明有底水补充, 能量充沛 突破型 
单位时间内含水率呈短暂下降的震荡上升趋势，表明水体能量占主导, 整体呈上升趋势 反转型 
单位时间内含水率下降速度变缓, 连续出现反转加速上扬信号，表明有新的水体能量沟通, 含水率明显回升 
下载: 导出CSV
表 3 LSTM不同神经元节点数量的准确率
Table 3. Accuracy of the number of different neurons in LSTM
节点数量 8 16 32 64 AC/% 86.84 97.36 93.31 92.11
下载: 导出CSV
表 4 LSTM不同隐藏层层数的准确率
Table 4. Accuracy of different hidden layers of LSTM
隐藏层层数 1 2 3 AC/% 97.36 94.73 92.10
下载: 导出CSV
表 5 LSTM不同学习率的准确率
Table 5. Accuracy of different learning rates of LSTM
学习率 0.001 0.005 0.01 0.015 AC/% 92.11 97.36 94.74 89.47
下载: 导出CSV
表 6 LSTM模型参数表
Table 6. Parameters of LSTM model
参数 取值 损失函数 交叉熵函数 优化算法 Adam 输出层激活函数 softmax 学习率 0.005 隐藏层层数 1 神经元节点数 16 迭代次数 20 000
下载: 导出CSV
-
[1] 康玉柱. 中国古生代碳酸盐岩古岩溶储集特征与油气分布[J]. 天然气工业, 2008, 28(6): 1-12. doi: 10.3787/j.issn.1000-0976.2008.06.001Kang Y Z. Characteristics and distribution laws of paleokarst hydrocarbon reservoirs in Palaeozoic carbonate formations in China[J]. Natural Gas Industry, 2008, 28(6): 1-12(in Chinese with English abstract). doi: 10.3787/j.issn.1000-0976.2008.06.001 [2] 王睿思. 塔河油田缝洞型油藏见水时间及含水变化规律预测[D]. 成都: 西南石油大学, 2012.Wang R S. Prediction of water breakthrough time and water cut variation of fracture-vuggy reservoir in Tahe Oilfield[D]. Chengdu: Southwest Petroleum University, 2012(in Chinese with English abstract). [3] 崔书岳, 康志江, 邸元. 基于多相流模型的缝洞型油藏数值模拟软件研制与应用[J]. 地质科技情报, 2019, 38(5): 97-104. https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ201905010.htmCui S Y, Kang Z J, Di Y. Development and application of numerical simulation software platform for fractured-cave reservoir based on multiphase flow model[J]. Geological Science and Technology Information, 2019, 38(5): 97-104(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ201905010.htm [4] 张冬梅, 林梓航, 康志江, 等, 基于EEMD高斯过程自回归模型的缝洞型油藏开发动态指标预测[J]. 地质科技情报, 2019, 38(2): 257-264. https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ201903028.htmZhang D M, Lin Z H, Kang Z J, et al. Prediction of development dynamic indexs for fractured-vuggy carbonate reservoir based on EEMD and Gaussian process autoregression model[J]. Geological Science and Technology Information, 2019, 38(2): 257-264(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ201903028.htm [5] 刘志远, 冒海军, 鄢宇杰, 等. 塔河油田托甫台区块奥陶系一间房组裂缝分布特征[J]. 地质科技情报, 2019, 38(5): 64-70. https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ201905006.htmLiu Z Y, Mao H J, Yan Y J, et al. Distribution characteristics of fractures in the Ordovician Yijianfang Formation in Tuofutai area, Tahe Oilfield[J]. Geological Science and Technology Information, 2019, 38(5): 64-70(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ201905006.htm [6] 罗娟, 鲁新便, 巫波, 等. 塔河油田缝洞型油藏高产油井见水预警评价技术[J]. 石油勘探与开发, 2013, 40(4): 468-473. https://www.cnki.com.cn/Article/CJFDTOTAL-SKYK201304011.htmLuo J, Lu X B, Wu B, et al. A water breakthrough warning system of high-yield wells in fractured-vuggy reservoirs in Tahe Oilfield[J]. Petroleum Exploration and Development, 2013, 40(4): 468-473(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-SKYK201304011.htm [7] 杨占红, 龙喜彬, 巫波, 等. 塔河油田缝洞型油藏高产井见水特征及水淹预警机制的建立[J]. 西部探矿工程, 2014, 26(3): 55-58. doi: 10.3969/j.issn.1004-5716.2014.03.020Yang Z H, Long X B, Wu B, et al. Water breakthrough characteristics of high-yield wells and establishment of flooding warning mechanism in fractured-vuggy reservoir in Tahe Oilfield[J]. West-China Exploration Engineering, 2014, 26(3): 55-58(in Chinese with English abstract). doi: 10.3969/j.issn.1004-5716.2014.03.020 [8] 李小波, 荣元帅, 龙喜彬, 等. 缝洞型油藏强边、底水窜进油井特征及机理研究[J]. 西南石油大学学报: 自然科学版, 2015, 37(1): 135-140. https://www.cnki.com.cn/Article/CJFDTOTAL-XNSY201501019.htmLi X B, Rong Y S, Long X B, et al. Characteristic and mechanism research of water flooding of strong edge bottom aquifer fractured-vuggy reservoir[J]. Journal of Southwest Petroleum University: Science & Technology Edition, 2015, 37(1): 135-140(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-XNSY201501019.htm [9] 邓明, 张国枢, 陈清华. 基于瓦斯涌出时间序列的煤与瓦斯突出预报[J]. 煤炭学报, 2010, 35(2): 260-263. https://www.cnki.com.cn/Article/CJFDTOTAL-MTXB201002028.htmDeng M, Zhang G S, Chen Q H. Forecast of coal and gas outburst based on time series of gas concentration[J]. Journal of China Coal Society, 2010, 35(2): 260-263(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-MTXB201002028.htm [10] 梁伟. 基于K线原理的矿井瓦斯异常诊断及预警研究[D]. 北京: 中国矿业大学, 2017.Liang W. Research on diagnosis and early warning of coal mine gas abnormality based on K line principle[D]. Beijing: China University of Mining and Technology, 2017(in Chinese with English abstract). [11] 魏连江, 胡青伟, 梁伟, 等. 基于K线图理论的瓦斯异常模式诊断研究[J]. 煤矿安全, 2019, 50(6): 24-27, 31. https://www.cnki.com.cn/Article/CJFDTOTAL-MKAQ201906006.htmWei L J, Hu Q W, Liang W, et al. Research on gas abnormal mode diagnosis based on K-line theory[J]. Safety in Coal Mines, 2019, 50(6): 24-27, 31(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-MKAQ201906006.htm [12] Hochreiter S, Schmidhuber J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780. doi: 10.1162/neco.1997.9.8.1735 [13] Graves A. Long short-term memory[M]. Berlin: Springer, 2012: 1735-1780. [14] Bengio Y. Learning long-term dependencies with gradient descent is difficult[J]. IEEE Transactions on Neural Networks, 1994, 5(2): 157-166. doi: 10.1109/72.279181 -
下载:
点击查看大图
计量
- 文章访问数: 603
- PDF下载量: 269
- 被引次数: 0
