Method of Predicting Drilling Fluid Rheology Based on Generative Adversarial Networks in Digital Twin Environment
-
摘要: 为了解决实验室中人工测量钻井液流变性能效率低、成本高、稳定性差的问题，提出了数字孪生环境下的基于生成对抗网络的钻井液流变性能预测方法。首先，根据数字孪生五维模型构建了钻井液配制与测量系统的孪生模型，物理配测系统中的传感器等信息采集器会收集钻井液流变性能测试实验中的物理实况数据，整合钻井液配方信息和实验测量结果后传输至虚拟空间，建立钻井液流变性能预测数据库；然后，利用改进的生成对抗网络算法，构建钻井液流变性能预测模型。从数据库中抽取钻井液历史孪生数据作为数据集对模型进行训练，得到最佳拟合模型，通过钻井液流变性能预测实验验证模型的预测能力。最终结果表明，模型预测值和真实值之间的相关系数R超过0.96，平均绝对百分比误差 AAPE 不高于4.1%，模型具有较高的预测精度，能够完成钻井液流变性能预测任务。Abstract: A method of predicting drilling fluid rheology based on genitive adversarial network in digital twin environment has been developed to deal with problems in laboratory measurement of drilling fluid rheology manually, such as low efficiency, high cost and poor stability etc. First, a twin model for drilling fluid formulation and measurement system is constructed in accordance with digital twin five-dimensional model. Information collectors such as sensors in the physical measurement system can collect the live measured data of drilling fluid properties, and the composition of the drilling fluid and the measured drilling fluid properties are integrated and sent to a virtual space to construct a database for drilling fluid property prediction. Second, using the improved generative adversarial network algorithm, a drilling fluid rheology prediction model is constructed. Historical twin data of the drilling fluid are then extracted from the database and are used as a dataset to train the model, and a best-fitting model is thus obtained. The prediction ability of the model is verified through the prediction experiment of drilling fluid rheology. Use of the model shows that the correlation coefficient R between the predicted values and the true values exceeds 0.96, and the mean absolute percentage error (AAPE) is not higher than 4.1%, indicating that the model has higher prediction accuracy, and is able to accomplish the predciiton of drilling fluid rheology.
-
Key words:
- Digital twin /
- Generative adversarial network /
- Drilling fluid /
- Property prediction
-
表 1 模型预测值与真实值相关度指标汇总
统计指标 φ600 φ300 φ200 φ100 R 0.96 0.98 0.95 0.96 AAPE/% 3.10 2.90 3.50 3.10 MAE 1.36 0.73 0.70 0.35 RMSE 1.60 0.89 0.80 0.42 R2 0.93 0.96 0.97 0.10 
下载: 导出CSV
表 2 配方真实数据与模型预测数据的绝对误差
序号 表观黏度真实值/
mPa·s表观黏度预测值/
mPa·s绝对误差 塑性黏度真实值/
mPa·s塑性黏度预测值/
mPa·s绝对误差 1 42.13 45.56 3.43 34.10 36.60 2.50 2 47.36 48.45 1.09 42.54 44.68 2.14 3 43.65 41.03 2.62 33.55 32.56 0.99 4 42.54 42.49 0.05 36.51 34.76 1.75 5 36.19 36.76 0.57 29.25 27.80 1.45 6 43.22 43.15 0.07 29.72 33.15 3.43 7 36.98 37.20 0.22 30.66 31.30 0.64 8 28.67 28.18 0.49 25.55 25.75 0.20 9 44.08 44.56 0.48 37.57 35.29 2.28 10 51.35 49.69 1.66 45.67 42.98 2.69 11 46.63 48.71 2.08 38.98 40.01 1.03 12 51.81 49.75 2.06 46.86 44.45 2.41 13 28.82 27.84 0.98 25.65 25.55 0.10 14 43.96 40.67 3.29 35.67 33.15 2.52 15 50.46 48.32 2.14 45.08 41.52 3.56 16 35.00 37.78 2.78 27.51 27.27 0.24 17 44.36 41.67 2.69 37.03 34.66 2.37 18 50.51 48.62 1.89 45.57 44.55 1.02 19 36.61 37.55 0.94 27.92 27.75 0.17 20 41.07 45.23 4.16 32.89 35.76 2.87
下载: 导出CSV
-
[1] CHEN L. Optimization design of drilling fluid chemical formula based on artificial intelligence[J]. Computational Intelligence and Neuroscience, 2022, 2022:5465816. [2] ROOKI R. Application of general regression neural network (GRNN) for indirect measuring pressure loss of Herschel-Bulkley drilling fluids in oil drilling[J]. Measurement, 2016, 85:184-191. doi: 10.1016/j.measurement.2016.02.037 [3] ABDELAAL A, IBRAHIM A, ELKATATNY S. Rheological properties prediction of flat rheology drilling fluids[C]//Paper presented at the 56th U. S. Rock Mechanics/Geomechanics Symposium. Santa Fe, New Mexico, USA, 2022: ARMA-2022-0822. [4] GAUTAM S, GURIA C, GOPE L. Prediction of high-pressure/high-temperature rheological properties of drilling fluids from the viscosity data measured on a coaxial cylinder viscometer[J]. SPE Journal, 2021, 26(5):2527-2548. doi: 10.2118/206714-PA [5] FAN H, ZHOU H, MENG X, et al. Accurate prediction model for rheological properties of drilling fluids at high temperature and high pressure conditions[C]//Paper presented at the SPE/IATMI Asia Pacific Oil & Gas Conference and Exhibition. Nusa Dua, Bali, Indonesia, 2015: SPE-176263-MS. [6] GOLSEFATAN A, SHAHBAZI K. Predicting performance of SiO2 nanoparticles on filtration volume using reliable approaches: application in water-based drilling fluids[J]. Energy Sources Part A: Recovery Utilization and Environmental Effects, 2021, 43(23):3216-3225. doi: 10.1080/15567036.2019.1639854 [7] ELKATATNY S, TARIQ Z, MAHMOUD M. Real time prediction of drilling fluid rheological properties using Artificial Neural Networks visible mathematical model (white box)[J]. Journal of Petroleum Science and Engineering, 2016, 146:1202-1210. doi: 10.1016/j.petrol.2016.08.021 [8] 周号博. 基于黏度计读值预测的高温高压流变性预测方法[J]. 钻井液与完井液,2019,36(3):325-332. doi: 10.3969/j.issn.1001-5620.2019.03.011ZHOU Haobo. A method of predicting high temperature high pressure rheological property based on viscometer readings[J]. Drilling Fluid & Completion Fluid, 2019, 36(3):325-332. doi: 10.3969/j.issn.1001-5620.2019.03.011 [9] 刘保双, 王忠杰, 马云谦, 等. 钻井液流变性在线检测新方法[J]. 钻井液与完井液,2016,33(4):56-59.LIU Baoshuang, WANG Zhongjie, MA Yunqian, et al. New method of online measurement of drilling fluid rheology[J]. Drilling Fluid & Completion Fluid, 2016, 33(4):56-59. [10] YANG C, WANG Z H, ZHENG L H, et al. Predicting equivalent static density of fuzzy ball drilling fluid by BP artificial neutral network[J]. Advances in Materials Science and Engineering, 2015, 2015(1):340721. [11] TOMIWA O, OLUWATOSIN R, TEMILOLUWA O, et al.Improved water based mud using solanum tuberosum formulated biopolymer and application of artificial neural network in predicting mud rheological properties [C]//Paper presented at the SPE Nigeria Annual International Conference and Exhibition. Lagos, Nigeria, 2019: SPE-198861-MS. [12] ZHANG D, DUAN L C, XU Y, et al. An ANN-based soft-sensor to estimate the sand content of drilling fluid[C]//2022 IEEE 6th Advanced Information Technology, Electronic and Automation Control Conference (IAEAC). Beijing, China, 2022: 1598-1602. [13] 陶飞, 刘蔚然, 张萌, 等. 数字孪生五维模型及十大领域应用[J]. 计算机集成制造系统,2019,25(1):1-18.TAO Fei, LIU Weiran, ZHANG Meng, et al. Five-dimension digital twin model and its ten applications[J]. Computer Integrated Manufacturing Systems, 2019, 25(1):1-18. [14] KANFAR R, HALABI L E, HALL T, et al. Well placement optimization for avoiding caves using GANs and POMDPs[C]//Paper presented at the International Petroleum Technology Conference. Dhahran, Saudi Arabia, 2024: IPTC-24209-MS. [15] SHARIFINASAB M H, EMAMI NIRI M, MASROOR M. Developing GAN-boosted Artificial Neural Networks to model the rate of drilling bit penetration[J]. Applied Soft Computing, 2023, 136:110067. doi: 10.1016/j.asoc.2023.110067 [16] AKHTAR M S, ALI A, CHAUDHURI S S. Mobile-UNet GAN: a single-image dehazing model[J]. Signal Image and Video Processing, 2024, 18(1):275-283. doi: 10.1007/s11760-023-02752-3 [17] SIDDIG O M, ELKATATNY S. Utilizing drilling data and machine learning in Real-Time prediction of poisson's ratio[C]//Paper presented at the Middle East Oil, Gas and Geosciences Show. Manama, Bahrain, 2023: SPE-213694-MS. -
点击查看大图
图(8) / 表(2)
计量
- 文章访问数: 12
- HTML全文浏览量: 6
- PDF下载量: 3
- 被引次数: 0
