Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
158 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
45 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Physics-Informed Neural Networks with Skip Connections for Modeling and Control of Gas-Lifted Oil Wells (2403.02289v1)

Published 4 Mar 2024 in cs.LG

Abstract: Neural networks, while powerful, often lack interpretability. Physics-Informed Neural Networks (PINNs) address this limitation by incorporating physics laws into the loss function, making them applicable to solving Ordinary Differential Equations (ODEs) and Partial Differential Equations (PDEs). The recently introduced PINC framework extends PINNs to control applications, allowing for open-ended long-range prediction and control of dynamic systems. In this work, we enhance PINC for modeling highly nonlinear systems such as gas-lifted oil wells. By introducing skip connections in the PINC network and refining certain terms in the ODE, we achieve more accurate gradients during training, resulting in an effective modeling process for the oil well system. Our proposed improved PINC demonstrates superior performance, reducing the validation prediction error by an average of 67% in the oil well application and significantly enhancing gradient flow through the network layers, increasing its magnitude by four orders of magnitude compared to the original PINC. Furthermore, experiments showcase the efficacy of Model Predictive Control (MPC) in regulating the bottom-hole pressure of the oil well using the improved PINC model, even in the presence of noisy measurements.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (27)
  1. Physics-informed neural nets for control of dynamical systems. arXiv preprint arXiv:2104.02556 .
  2. Model Predictive Control. Springer Science & Business Media.
  3. Physics informed extreme learning machine (PIELM)–a rapid method for the numerical solution of partial differential equations. Neurocomputing 391, 96–118. doi:10.1016/j.neucom.2019.12.099.
  4. Model predictive control in aerospace systems: Current state and opportunities. Journal of Guidance, Control, and Dynamics 40, 1541–1566. doi:10.2514/1.G002507.
  5. Practical Methods of Optimization. John Wiley & Sons.
  6. Physics informed neural networks for control oriented thermal modeling of buildings. Applied Energy 314, 118852. doi:10.1016/j.apenergy.2022.118852.
  7. Selected Topics on Constrained and Nonlinear Control. STU/NTNU.
  8. Control structure design for stabilizing unstable gas-lift oil wells. IFAC Proceedings Volumes 45, 93–100. doi:10.3182/20120710-4-SG-2026.00110.
  9. Echo State Networks for Online Learning Control and MPC of Unknown Dynamic Systems: Applications in the Control of Oil Wells. Master’s thesis. Universidade Federal de Santa Catarina.
  10. Echo state networks for practical nonlinear model predictive control of unknown dynamic systems. IEEE Transactions on Neural Networks and Learning Systems 33, 2615–2629. doi:10.1109/TNNLS.2021.3136357.
  11. Physics-informed machine learning. Nature Reviews Physics 3, 422–440. doi:10.1038/s42254-021-00314-5.
  12. Physics-informed Dyna-style model-based deep reinforcement learning for dynamic control. Proceedings of the Royal Society A 477, 20210618. doi:10.1098/rspa.2021.0618.
  13. Short-term steady-state production optimization of offshore oil platforms: wells with dual completion (gas-lift and esp) and flow assurance. TOP 30. doi:10.1007/s11750-021-00604-28.
  14. Nonlinear model predictive control for trajectory tracking of nonholonomic mobile robots: A modified approach. International Journal of Advanced Robotic Systems 15. doi:10.1177/1729881418760461.
  15. Physics-informed neural networks for modeling water flows in a river channel. IEEE Transactions on Artificial Networks , 1–15. doi:10.1109/TAI.2022.3200028.
  16. Physics-informed neural networks-based model predictive control for multi-link manipulators. IFAC-PapersOnLine 55, 331–336. doi:10.1016/j.ifacol.2022.09.117.
  17. Reconstruction of hydrofoil cavitation flow based on the chain-style physics-informed neural network. Engineering Applications of Artificial Intelligence 119, 105724. doi:10.1016/j.engappai.2022.105724.
  18. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. Journal of Computational Physics 378, 686–707. doi:0.1016/j.jcp.2018.10.045.
  19. Nonlinear model predictive control using successive linearization. Transactions of the Society of Instrument and Control Engineers 38, 61–66. doi:10.9746/sicetr1965.38.61.
  20. Application of a parallel physics-informed neural network to solve the multi-body dynamic equations for full-scale train collisions. Applied Soft Computing 142, 110328. doi:10.1016/j.asoc.2023.110328.
  21. Optimizing the optimizer for data driven deep neural networks and physics informed neural networks. arXiv preprint arXiv:2205.07430 .
  22. On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming. Mathematical Programming 106, 25–57. doi:10.1007/s10107-004-0559-y.
  23. Understanding and mitigating gradient flow pathologies in physics-informed neural networks. SIAM Journal on Scientific Computing 43, A3055–A3081. doi:10.1137/20M1318043.
  24. Navier–stokes generative adversarial network: a physics-informed deep learning model for fluid flow generation. Neural Computing and Applications 34, 11539–11552. doi:10.1007/s00521-022-07042-6.
  25. Self-adaptive loss balanced physics-informed neural networks. Neurocomputing 496, 11–34. doi:10.1016/j.neucom.2022.05.015.
  26. Successive linearization based model predictive control of variable stiffness actuated robots, in: IEEE International Conference on Advanced Intelligent Mechatronics (AIM), pp. 1774–1779. doi:10.1109/AIM.2017.8014275.
  27. Physics-informed deep learning: A promising technique for system reliability assessment. Applied Soft Computing 126, 109217. doi:10.1016/j.asoc.2022.109217.
Citations (1)

Summary

We haven't generated a summary for this paper yet.