# Nonlinear Autoregressive Neural Networks to Predict Hydraulic Fracturing Fluid Leakage into Shallow Groundwater

^{1}

^{2}

^{3}

^{*}

## Abstract

**:**

^{2}) between measured and predicted values is 0.923 and the mean squared error (MSE) is 4.2 × 10

^{−4}, and the values of R

^{2}= 0.944 and MSE = 2.4 × 10

^{−4}were obtained for the NAR-BR model. The results indicate the robustness and compatibility of NAR-LM and NAR-BR models in predicting fracturing fluid flow rate to shallow aquifers. This study shows that NAR neural networks can be useful and hold considerable potential for assessing the groundwater impacts of unconventional gas development.

## 1. Introduction

## 2. Materials and Methods

#### 2.1. Conceptual Model for Fracturing Fluid Migration along an Abandoned Well

^{3}of fracturing fluid [13] is injected into the domain during 2.5 days. Then, a 5-day shut-in period takes place. Afterwards, the production periods are modeled with outflow rates of 16 and 3.2 m

^{3}/day for an interrupted period of 2 and 13 years, respectively [13]. In the abandonment stage, the production is stopped. The fluid flow along the abandoned well and at the overburden aquifer interface is monitored. Note that we used data reported in Brownlow et al. [13], which are based on available industry data for the Eagle Ford Shale play in south Texas.

#### 2.2. Data Preparation and Analysis

#### 2.3. NAR Model

#### 2.4. Training Algorithms

#### 2.4.1. Levenberg–Marquardt

#### 2.4.2. Bayesian Regularization

#### 2.5. Network Architecture

#### 2.6. Performance Evaluation

^{2}) and mean squared errors (MSE). R

^{2}is a linear regression used to analyze the best fit between the measured values and model’s predicted values, given by:

## 3. Results and Discussion

^{2}for the training phase (Table 2) show that both models reached the best fitting performances in terms of evaluation criteria. For instance, R

^{2}values of higher than 0.99 indicated a strong correlation between the measured values and fitting values. According to Figure 5 and Table 2, we conclude that a sample of 446 data is adequate to train the NAR models.

^{2}values higher than 0.9 for the testing phase indicated the satisfactory performance of both developed models. For further comparison of evaluation indices, MSE of NAR-LM and NAR-BAR were 4.2 × 10

^{−4}and 2.4 × 10

^{−4}respectively, which proved the strong ability of the developed models in predicting the nonlinear behavior of fracturing fluid flow to the aquifer. The results suggest that NAR-BR model has a slightly better prediction performance compared with that of the NAR-LM model in terms of larger R

^{2}and smaller MSE.

^{2}value of the NAR-BR model is slightly higher.

^{−5}at epoch 9 for the NAR-LM model, while the NAR-BR model experiences the best performance (MSE = 1.3 × 10

^{−5}) for the training phase at epoch 16.

^{−3}m

^{3}/year) for the entire time period, indicating that the NAR-LM and NAR-BR models were able to efficiently predict fracturing fluid flow rate to the aquifer over the simulation period. Comparing the response lines and error plots of this figure further indicated the superior performance of the NAR-BR model.

## 4. Conclusions

## Author Contributions

## Funding

## Acknowledgments

## Conflicts of Interest

## References

- Cook, T.; Perrin, J.; Wagener, D.V. Hydraulically Fractured Horizontal Wells Account for Most New Oil and Natural Gas Wells. Available online: https://www.eia.gov/todayinenergy/detail.php?id=34732 (accessed on 30 January 2018).
- Tatomir, A.; McDermott, C.; Bensabat, J.; Class, H.; Edlmann, K.; Taherdangkoo, R.; Sauter, M. Conceptual model development using a generic Features, Events, and Processes (FEP) database for assessing the potential impact of hydraulic fracturing on groundwater aquifers. Adv. Geosci.
**2018**, 45, 185–192. [Google Scholar] [CrossRef] [Green Version] - Vengosh, A.; Jackson, R.; Warner, N. A critical review of the risks to water resources from unconventional shale gas development and hydraulic fracturing in the United States. Sci. Technol.
**2014**, 48, 8334–8348. [Google Scholar] [CrossRef] [PubMed] - Sauter, M.; Brosig, K.; Lange, T.; Jahnke, W.; Helmig, R.; Kissinger, A.; Heitfeld, M.; Kluenker, J.; Schetelig, K. Risiken im Geologischen System bei der Fracking-Technologie Abschätzung der Auswirkungen auf Grundwasservorkommen. Wasser und Abfall
**2012**, 14, 16. [Google Scholar] - Rice, A.K.; Lackey, G.; Proctor, J.; Singha, K. Groundwater-quality hazards of methane leakage from hydrocarbon wells: A review of observational and numerical studies and four testable hypotheses. Wiley Interdiscip. Rev. Water
**2018**, 5, e1283. [Google Scholar] [CrossRef] - Lange, T.; Sauter, M.; Heitfeld, M.; Schetelig, K.; Brosig, K.; Jahnke, W.; Kissinger, A.; Helmig, R.; Ebigbo, A.; Class, H. Hydraulic fracturing in unconventional gas reservoirs: Risks in the geological system part 1. Environ. Earth Sci.
**2013**, 70, 3839–3853. [Google Scholar] [CrossRef] - Gassiat, C.; Gleeson, T.; Lefebvre, R.; McKenzie, J. Hydraulic fracturing in faulted sedimentary basins: Numerical simulation of potential contamination of shallow aquifers over long time scales. Water Resour. Res.
**2013**, 49, 8310–8327. [Google Scholar] [CrossRef] [Green Version] - Myers, T. Potential Contaminant Pathways from Hydraulically Fractured Shale to Aquifers. Groundwater
**2012**, 50, 872–882. [Google Scholar] [CrossRef] [PubMed] - Nowamooz, A.; Lemieux, J.-M.; Molson, J.; Therrien, R. Numerical investigation of methane and formation fluid leakage along the casing of a decommissioned shale gas well. Water Resour. Res.
**2015**, 51, 4592–4622. [Google Scholar] [CrossRef] [Green Version] - Taherdangkoo, R.; Tatomir, A.; Taylor, R.; Sauter, M. Numerical investigations of upward migration of fracking fluid along a fault zone during and after stimulation. Energy Procedia
**2017**, 125, 126–135. [Google Scholar] [CrossRef] - Birdsell, D.T.; Rajaram, H.; Dempsey, D.; Viswanathan, H.S. Hydraulic fracturing fluid migration in the subsurface: A review and expanded modeling results. Water Resour. Res.
**2015**, 51, 7159–7188. [Google Scholar] [CrossRef] - Reagan, M.T.; Moridis, G.J.; Keen, N.D.; Johnson, J.N. Numerical simulation of the environmental impact of hydraulic fracturing of tight/shale gas reservoirs on near-surface groundwater: Background, base cases, shallow reservoirs, short-term gas, and water transport. Water Resour. Res.
**2015**, 4, 2543–2573. [Google Scholar] [CrossRef] [PubMed] - Brownlow, J.W.; James, S.C.; Yelderman, J.C. Influence of Hydraulic Fracturing on Overlying Aquifers in the Presence of Leaky Abandoned Wells. Groundwater
**2016**, 54, 781–792. [Google Scholar] [CrossRef] [PubMed] - King, G.E.; Rainbolt, M.F.; Swanson, C. Frac Hit Induced Production Losses: Evaluating Root Causes, Damage Location, Possible Prevention Methods and Success of Remedial Treatments. In Proceedings of the SPE Annual Technical Conference and Exhibition, San Antonio, TX, USA, 9–11 October 2017. [Google Scholar]
- Gasda, S.E.; Bachu, S.; Celia, M.A. Spatial characterization of the location of potentially leaky wells penetrating a deep saline aquifer in a mature sedimentary basin. Environ. Geol.
**2004**, 46, 707–720. [Google Scholar] [CrossRef] - Gasda, S.E.; Celia, M.A.; Wang, J.Z.; Duguid, A. Wellbore permeability estimates from vertical interference testing of existing wells. Energy Procedia
**2013**, 37, 5673–5680. [Google Scholar] [CrossRef] [Green Version] - Taherdangkoo, R.; Tatomir, A.; Anighoro, T.; Sauter, M. Modeling fate and transport of hydraulic fracturing fluid in the presence of abandoned wells. J. Contam. Hydrol.
**2019**, 221, 58–68. [Google Scholar] [CrossRef] [PubMed] - Flewelling, S.A.; Sharma, M. Constraints on upward migration of hydraulic fracturing fluid and brine. Groundwater
**2014**, 52, 9–19. [Google Scholar] [CrossRef] [Green Version] - Birdsell, D.T.; Rajaram, H.; Lackey, G. Imbibition of hydraulic fracturing fluids into partially saturated shale. Water Resour. Res.
**2015**, 51, 6787–6796. [Google Scholar] [CrossRef] - Edwards, R.W.J.; Doster, F.; Celia, M.A.; Bandilla, K.W. Numerical modeling of gas and water flow in shale gas formations with a focus on the fate of hydraulic fracturing fluid. Environ. Sci. Technol.
**2017**, 51, 13779–13787. [Google Scholar] [CrossRef] - Pfunt, H.; Houben, G.; Himmelsbach, T. Numerical modeling of fracking fluid migration through fault zones and fractures in the North German Basin. Hydrogeol. J.
**2016**, 24, 1343–1358. [Google Scholar] [CrossRef] - Osborn, S.G.; Vengosh, A.; Warner, N.R.; Jackson, R.B. Methane contamination of drinking water accompanying gas-well drilling and hydraulic fracturing. Proc. Natl. Acad. Sci. USA
**2011**, 108, 8172–8176. [Google Scholar] [CrossRef] [Green Version] - Kissinger, A.; Helmig, R.; Ebigbo, A.; Class, H.; Lange, T.; Sauter, M.; Heitfeld, M.; Klünker, J.; Jahnke, W. Hydraulic fracturing in unconventional gas reservoirs: Risks in the geological system, part 2: Modelling the transport of fracturing fluids, brine and methane. Environ. Earth Sci.
**2013**, 70, 3855–3873. [Google Scholar] [CrossRef] - Cao, V.; Schaffer, M.; Taherdangkoo, R.; Licha, T. Solute reactive tracers for hydrogeological applications: A short review and future prospects. Water
**2020**, 12, 653. [Google Scholar] [CrossRef] [Green Version] - Guzman, S.M.; Paz, J.O.; Tagert, M.L.M. The use of NARX neural networks to forecast daily groundwater levels. Water Resour. Manag.
**2017**, 31, 1591–1603. [Google Scholar] [CrossRef] - Loucks, D.P.; Van Beek, E. Water Resources Systems Planning and Management: An Introduction to Methods, Models and Applications. Available online: https://hydrologie.org/BIB/Publ_UNESCO/SR_999_E_2005.pdf (accessed on 15 March 2020).
- Kisi, O.; Cigizoglu, H.K. Comparison of different ANN techniques in river flow prediction. Civ. Eng. Environ. Syst.
**2007**, 24, 211–231. [Google Scholar] [CrossRef] - Olyaie, E.; Banejad, H.; Chau, K.W.; Melesse, A.M. A comparison of various artificial intelligence approaches performance for estimating suspended sediment load of river systems: A case study in United States. Environ. Monit. Assess.
**2015**, 187, 189. [Google Scholar] [CrossRef] - Zhang, Q.; Stanley, S.J. Forecasting raw-water quality parameters for the north Saskatchewan river by neural network modeling. Water Res.
**1997**, 31, 2340–2350. [Google Scholar] [CrossRef] - Maier, H.R.; Dandy, G.C. The Use of Artificial Neural Networks for the Prediction of Water Quality Parameters. Water Resour. Res.
**1996**, 32, 1013–1022. [Google Scholar] [CrossRef] - Wei, S.; Zuo, D.; Song, J. Improving prediction accuracy of river discharge time series using a Wavelet-NAR artificial neural network. J. Hydroinformatics
**2012**, 14, 974–991. [Google Scholar] [CrossRef] [Green Version] - Ruiz, L.; Cuéllar, M.; Calvo-Flores, M.; Jiménez, M. An application of non-linear autoregressive neural networks to predict energy consumption in public buildings. Energies
**2016**, 9, 684. [Google Scholar] [CrossRef] [Green Version] - Adamowski, J.; Fung Chan, H.; Prasher, S.O.; Ozga-Zielinski, B.; Sliusarieva, A. Comparison of multiple linear and nonlinear regression, autoregressive integrated moving average, artificial neural network, and wavelet artificial neural network methods for urban water demand forecasting in Montreal, Canada. Water Resour. Res.
**2012**, 48, 1–14. [Google Scholar] [CrossRef] - Neuzil, C.E. How permeable are clays and shales? Water Resour. Res.
**1994**, 30, 145–150. [Google Scholar] [CrossRef] - Edwards, R.W.; Celia, M.A. Shale gas well, hydraulic fracturing, and formation data to support modeling of gas and water flow in shale formations. Water Resour. Res.
**2018**, 54, 3196–3206. [Google Scholar] [CrossRef] - Freeze, R.A.; Cherry, J.A. Groundwater prentice-hall. Eaglewood Cliffs
**1979**, 176, 161–177. [Google Scholar] - Gleeson, T.; Smith, L.; Moosdorf, N.; Hartmann, J.; Dürr, H.H.; Manning, A.H.; van Beek, L.P.H.; Jellinek, A.M. Mapping permeability over the surface of the Earth. Geophys. Res. Lett.
**2011**, 38, L02401. [Google Scholar] [CrossRef] [Green Version] - Luijendijk, E.; Gleeson, T. How well can we predict permeability in sedimentary basins? Deriving and evaluating porosity-permeability equations for noncemented sand and clay mixtures. Geofluids
**2015**, 15, 67–83. [Google Scholar] [CrossRef] [Green Version] - EIA. World Shale Gas Resources: An Initial Assessment of 14 Regions Outside the United States; US Department of Energy: Washington, DC, USA, 2011. [Google Scholar]
- EIA. Review of Emerging Resources: US Shale Gas and Shale Oil Plays; US Department of Energy: Washington, DC, USA, 2011. [Google Scholar]
- Magri, F.; Bayer, U.; Jahnke, C.; Clausnitzer, V.; Diersch, H.J.; Fuhrman, J.; Möller, P.; Pekdeger, A.; Tesmer, M.; Voigt, H.J. Fluid-dynamics driving saline water in the North East German Basin. Int. J. Earth Sci.
**2005**, 94, 1056–1069. [Google Scholar] [CrossRef] - Kissinger, A.; Noack, V.; Knopf, S.; Konrad, W.; Scheer, D.; Class, H. Regional-scale brine migration along vertical pathways due to CO
_{2}injection--Part 2: A simulated case study in the North German Basin. Hydrol. Earth Syst. Sci.**2017**, 21, 2751–2775. [Google Scholar] [CrossRef] [Green Version] - Nogues, J.P.; Nordbotten, J.M.; Celia, M.A. Detecting leakage of brine or CO
_{2}through abandoned wells in a geological sequestration operation using pressure monitoring wells. Energy Procedia**2011**, 4, 3620–3627. [Google Scholar] [CrossRef] [Green Version] - Bai, M.; Reinicke, K. In Numerical Simulation of CO
_{2}Leakage through Abandoned Wells during CO_{2}Underground Storage, Proceedings of the Clean Energy Systems in the Subsurface: Production, Storage and Conversion, Goslar, Germany, 21–23 May 2013; Springer: Berlin/Heidelberg, Germany, 2013; pp. 197–210. [Google Scholar] - Pawar, R.J.; Watson, T.L.; Gable, C.W. Numerical simulation of CO
_{2}leakage through abandoned wells: Model for an abandoned site with observed gas migration in Alberta, Canada. Energy Procedia**2009**, 1, 3625–3632. [Google Scholar] [CrossRef] [Green Version] - Gasda, S.E.; Nordbotten, J.M.; Celia, M.A. Determining effective wellbore permeability from a field pressure test: A numerical analysis of detection limits. Environ. Geol.
**2008**, 54, 1207–1215. [Google Scholar] [CrossRef] - Nordbotten, J.M.; Celia, M.A.; Bachu, S.; Dahle, H.K. Semianalytical solution for CO
_{2}leakage through an abandoned well. Environ. Sci. Technol.**2005**, 39, 602–611. [Google Scholar] [CrossRef] [PubMed] - Nordbotten, J.M.; Celia, M.A.; Bachu, S. Analytical solutions for leakage rates through abandoned wells. Water Resour. Res.
**2004**, 40, W04204. [Google Scholar] [CrossRef] - Ebigbo, A.; Class, H.; Helmig, R. CO
_{2}leakage through an abandoned well: Problem-oriented benchmarks. Comput. Geosci.**2007**, 11, 103–115. [Google Scholar] [CrossRef] - Celia, M.A.; Bachu, S.; Nordbotten, J.M.; Gasda, S.E.; Dahle, H.K. Quantitative estimation of CO 2 leakage from geological storage: Analytical models, numerical models, and data needs. Greenh. Gas Control Technol. 7
**2005**, I, 663–671. [Google Scholar] - Da Costa Lopes, F.; Watanabe, E.H.; Rolim, L.G.B. A control-oriented model of a PEM fuel cell stack based on NARX and NOE neural networks. IEEE Trans. Ind. Electron.
**2015**, 62, 5155–5163. [Google Scholar] [CrossRef] - López, M.; Valero, S.; Senabre, C.; Aparicio, J.; Gabaldon, A. Application of SOM neural networks to short-term load forecasting: The Spanish electricity market case study. Electr. Power Syst. Res.
**2012**, 91, 18–27. [Google Scholar] [CrossRef] - Wang, K.W.; Deng, C.; Li, J.P.; Zhang, Y.Y.; Li, X.Y.; Wu, M.C. Hybrid methodology for tuberculosis incidence time-series forecasting based on ARIMA and a NAR neural network. Epidemiol. Infect.
**2017**, 145, 1118–1129. [Google Scholar] [CrossRef] [Green Version] - Maier, H.R.; Dandy, G.C. Neural networks for the prediction and forecasting of water resources variables: A review of modelling issues and applications. Environ. Model. Softw.
**2000**, 15, 101–124. [Google Scholar] [CrossRef] - Hagan, M.T.; Menhaj, M.B. Training Feedforward Networks with the Marquardt Algorithm. IEEE Trans. Neural Netwrok
**1994**, 5, 989–993. [Google Scholar] [CrossRef] - Marquardt, D.W. An Algorithm for Least-Squares Estimation of Nonlinear Parameters. J. Soc. Ind. Appl. Math.
**1963**, 11, 431–441. [Google Scholar] [CrossRef] - Sahoo, S.; Jha, M.K. Groundwater-level prediction using multiple linear regression and artificial neural network techniques: A comparative assessment. Hydrogeol. J.
**2013**, 21, 1865–1887. [Google Scholar] [CrossRef] - Bishop, C. Neural Networks for Pattern Recognition; Oxford University Press: Oxford, UK, 1995. [Google Scholar]
- MacKay, D.J.C. Bayesian Interpolation. Neural Comput.
**1992**, 4, 415–447. [Google Scholar] [CrossRef] - Burden, F.; Winkler, D. Bayesian regularization of neural networks. Methods Mol. Biol.
**2008**, 458, 25–44. [Google Scholar] [PubMed] - Doan, C.D.; Liong, S.Y. Generalization for multilayer neural network bayesian regularization or early stopping. In Proceedings of the Asia Pacific Association of Hydrology and Water Resources 2nd Conference, Singapore, Singapore, 5–8 July 2004; pp. 5–8. [Google Scholar]
- Foresee, F.; Hagan, M.T. Gauss-Newton approximation to bayesian learning. In Proceedings of the IEEE International Conference on Neural Networks–Conference Proceedings, Houston, TX, USA, 12 June 1997; pp. 1930–1935. [Google Scholar]
- MacKay, D.J.C. A practical bayesian framework for backpropagation networks. Neural Comput.
**1992**, 4, 448–472. [Google Scholar] [CrossRef]

**Figure 1.**The conceptual model used for simulations in Taherdangkoo et al. [17].

**Figure 2.**Fracturing fluid flow rate to the aquifer for the base-case model during the simulation time.

**Figure 6.**Performance of the NAR-LM model for the prediction of fracturing fluid flow rate to the aquifer: (

**a**) training and (

**b**) testing.

**Figure 7.**Performance of the NAR-BR model for the prediction of fracturing fluid flow rate to the aquifer: (

**a**) training and (

**b**) testing.

**Figure 8.**Convergence plots of the NAR models in terms of number of epochs: (

**a**) NAR-LM and (

**b**) NAR-BR.

**Figure 9.**Response of (

**a**) NAR-LM and (

**b**) NAR-BR models in predicting fracturing fluid flow rate to the shallow aquifer. The top panels show the modeled fluid flow rate presented in Figure 2. The bottom panels display the model’s performance error in the training, validation, and testing phases.

**Table 1.**Parameters used in base-case model and sensitivity analysis simulations [17].

Parameter | Unit | Base-Case Value | Min. | Max. | Source |
---|---|---|---|---|---|

Shale permeability | m^{2} | 1 × 10^{−19} | 1 × 10^{−21} | 1 × 10^{−18} | [12,13,34] |

Shale porosity | 0.01 | 0.01 | 0.05 | [7,34,35,36] | |

Overburden permeability | m^{2} | Depth-dependent | 1 × 10^{−17} | 1 × 10^{−15} | [7,37,38] |

Overburden thickness | m | 1600 | 900 | 2900 | [7,39,40] |

Salinity gradient | g/lm | 0.15 | 0.1 | 0.2 | [41,42] |

Fracturing fluid volume | m^{3} | 11,365 | 11,000 | 15,000 | [8,11,13] |

Abandoned well permeability | m^{2} | 1 × 10^{−12} | 1 × 10^{−17} | 1 × 10^{−12} | [43,44,45,46,47,48,49,50] |

Distance of fracture plane to well | m | 0 | 0 | 15 |

Statistical Parameter | NAR-LM | NAR-BR | |||
---|---|---|---|---|---|

Training | Validation | Testing | Training | Testing | |

R^{2} | 0.998 | 0.996 | 0.923 | 0.996 | 0.944 |

MSE | 1.07 × 10^{−5} | 1.2 × 10^{−5} | 4.2 × 10^{−4} | 1.3 × 10^{−5} | 2.4 × 10^{−4} |

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Taherdangkoo, R.; Tatomir, A.; Taherdangkoo, M.; Qiu, P.; Sauter, M.
Nonlinear Autoregressive Neural Networks to Predict Hydraulic Fracturing Fluid Leakage into Shallow Groundwater. *Water* **2020**, *12*, 841.
https://doi.org/10.3390/w12030841

**AMA Style**

Taherdangkoo R, Tatomir A, Taherdangkoo M, Qiu P, Sauter M.
Nonlinear Autoregressive Neural Networks to Predict Hydraulic Fracturing Fluid Leakage into Shallow Groundwater. *Water*. 2020; 12(3):841.
https://doi.org/10.3390/w12030841

**Chicago/Turabian Style**

Taherdangkoo, Reza, Alexandru Tatomir, Mohammad Taherdangkoo, Pengxiang Qiu, and Martin Sauter.
2020. "Nonlinear Autoregressive Neural Networks to Predict Hydraulic Fracturing Fluid Leakage into Shallow Groundwater" *Water* 12, no. 3: 841.
https://doi.org/10.3390/w12030841