Control of Human-Induced Seismicity in Underground Reservoirs Governed by a Nonlinear PDE-ODE System

summary

The gist

Induced seismicity caused by fluid extraction or injection in underground reservoirs is a major challenge for safe energy production and storage The paper presents a robust output-feedback controller…

In short

The paper presents a robust output-feedback controller for mitigating induced seismicity from fluid extraction or injection in underground reservoirs. It uses a coupled 3D PDE-ODE model and a MIMO Super-Twisting algorithm to ensure that regional pressure and seismicity rates track desired targets while handling uncertainties.

Key concepts

Induced Seismicity Mitigation
This is the main goal: preventing earthquakes caused by injecting or extracting fluids underground. The system uses a controller to adjust fluid flow in real-time based on measurements to keep seismic activity within safe limits, which is crucial for safe energy production.
MIMO Super-Twisting Algorithm
This is the specific control strategy used. It's a nonlinear method designed to create a continuous control signal that is very robust against uncertainties in the system's parameters. It ensures the system tracks its desired targets effectively, even when things are not perfectly known.
PDE-ODE Model
The physical system being studied is modeled by two types of equations: a Partial Differential Equation (PDE) describing fluid diffusion within the reservoir and an Ordinary Differential Equation (ODE) describing the resulting nonlinear dynamics. This combined model captures how pressure changes spread through the rock and how seismicity evolves over time.

Terminology used across episodes

This episode discusses

The paper

Control of Human-Induced Seismicity in Underground Reservoirs Governed by a Nonlinear PDE-ODE System · Read on arXiv

Diego Gutiérez-Oribio, Ioannis Stefanou

IMSIA · CNRS · EDF · ENSTA Paris · Institut Polytechnique de Paris

Induced seismicity caused by fluid extraction or injection in underground reservoirs is a major challenge for safe energy production and storage. This paper presents a robust output-feedback controller for induced seismicity mitigation in geological reservoirs described by a coupled 3D PDE-ODE model. The controller is nonlinear and robust (MIMO Super-Twisting design), producing a continuous control signal and requiring minimal model information, while accommodating parameter uncertainties and spatial heterogeneity. Two operational outputs are regulated simultaneously: regional pressures and seismicity rates computed over reservoir sub-regions. Closed-loop properties are established via explicit bounds on the solution and its time derivative for both the infinite-dimensional dynamics and the nonlinear ODE system, yielding finite-time or exponential convergence of the tracking errors. The method is evaluated on the Groningen gas-field case study in two scenarios: gas production while not exceeding the intrinsic seismicity of the region, and combined production with CO 2 injection toward net-zero carbon operation. Simulations demonstrate accurate tracking of pressure and seismicity targets across regions under significant parameter uncertainty, supporting safer reservoir operation while preserving production objectives.

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: We're looking at the paper: "Control of Human-Induced Seismicity in Underground Reservoirs Governed by a Nonlinear PDE-ODE System".

Dev: Here's the gist of the paper. Induced seismicity caused by fluid extraction or injection in underground reservoirs is a major challenge for safe energy production and storage The paper presents a robust output-feedback controller for induced seismicity mitigation in geological reservoirs described by a coupled 3D PDE-ODE model This nonlinear and robust (MIMO Super-Twisting design) controller produces a continuous control signal and requires minimal model information, while accommodating parameter uncertainties and spatial heterogeneity Closed-loop properties are established via explicit bounds on the solution and its time derivative for both the infinite-dimensional dynamics and the nonlinear ODE system, yielding finite-time or exponential convergence of the tracking errors The method is evaluated on the Groningen gas-field case study in two scenarios: gas production while not exceeding the intrinsic seismicity of the region, and combined production with CO2 injection toward net-zero carbon operation System Modeling and Objectives The system studied is governed by a cascade model involving a 3D diffusion PDE and a logistic-like nonlinear ODE The objective of the work is to design the control input Q(t) of system (1) so that region-averaged pressure and seismicity-rate outputs track prescribed references The outputs are defined as regional averages of pressure yui(t) and seismicity rate yRi(t) The control design will be performed under several assumptions, including bounds on the actuation of the plant and constraints on the reference trajectories.

Rosa: Control Strategy Design The proposed robust output feedback controller is a Multi-Input Multi-Output (MIMO) Super-Twisting Algorithm This algorithm ensures robustness to uncertainties while generating a continuous control signal suitable for real actuators The error dynamics are formulated in matrix form as σ˙(t) = Ψ(t) + B(t)Q(t) The controller is defined as Q(t) = B+0h−K1 ⌈σ(t)1−l + ν(t), where ν˙(t) = −K2 ⌈σ(t)1+l/1−l The control signal generated is always continuous despite having a possible discontinuous integral term when l = −1.

Dev: Validation and Results The effectiveness of the controller is demonstrated through simulations conducted on a validated model of the Groningen gas reservoir In Scenario 1, control law (21) was implemented to regulate average pressure over five regions while controlling the average seismicity rate in the complementary region In Scenario 2, an additional constraint on the input Q(t) is incorporated by including CO2 injection to support carbon-neutral operation The results show that the control successfully drives all of the pressure outputs to their smooth references and maintains seismicity rates close to the reference R∗.

Rosa: Theoretical Guarantees The theoretical analysis establishes that there exist matrices K1, K2 defined as K1 = LK¯ 1, K2 = L 2K¯ 2, where L > 0 is sufficiently large The origin of system (16) is proven to be finite-time stable for l = −1 For l ∈ (−1, 0], the system exhibits finite-time input-to-state stability with respect to Ψ(¯ t) For l = 0, the system is exponentially input-to-state stable with respect to Ψ(¯ t) The analysis also proves that the set of points where the solution rate is unbounded must have zero measure.

Title and authors: Rosa: Let’s talk about the full title again, "Control of Human-Induced Seismicity in Underground Reservoirs Governed by a Nonlinear PDE-ODE System." It really captures the whole scope, doesn't it? It’s about controlling seismic activity caused by fluid extraction or injection in underground reservoirs.

Dev: I think the title tells you immediately that this isn't just a simple control problem; it involves coupling a PDE and an ODE system together. That complexity is what makes it challenging for standard methods to handle.

Taro: From my point of view, when you see that PDE-ODE structure, it suggests there’s this inherent dynamic interaction between how fluid moves through the rock—the diffusion part—and how the overall reservoir pressure changes over time—the ODE part.

Rosa: Right. And those kinds of systems are prevalent in modeling complex phenomena like heat transfer or electromagnetism, but here they're applied to a very specific, and very risky, application: preventing induced seismicity from energy operations.

Dev: The authors are Diego Gutiérrez-Oribio and Ioannis Stefanou. They’ve developed this robust output-feedback controller specifically for this coupled three dee PDE-ODE model <ref:2412.06327#pg1,coupled 3D PDE-ODE model>. That focus on the control aspect is what I find most important for the engineering side of things.

Taro: And when you look at their methods, they use a nonlinear and robust MIMO Super-Twisting design for the controller, which aims to generate a continuous control signal while handling parameter uncertainties and spatial heterogeneity.

Rosa: That’s right. The goal is to create a controller that’s not just theoretically stable on paper but can actually be implemented in the real world where you have those messy, uncertain conditions inside a reservoir.

Dev: What I want to focus on is that they manage this by generating a continuous control signal, which is great for real actuators, and it requires minimal model information. That’s a big deal because we usually need tons of data just to define the system dynamics initially.

Taro: Minimal model information sounds like a smart way to handle the complexity without having to build some huge, potentially inaccurate observer dynamics on top of the whole thing.

Rosa: Precisely. They rely only on measured outputs and nominal control information, which makes it much more practical than trying to perfectly observe an infinite-dimensional system. This approach is what they call elegant because it avoids those auxiliary observer dynamics that would otherwise be cumbersome for real applications.

Dev: So, this paper is presenting a method that bypasses the need for a separate observer system by relying directly on the measured outputs and nominal control information to achieve robust regulation of pressure and seismicity rates. That’s a significant simplification.

The paper's summary: Rosa: Let's go over what they actually achieved in their summary for this paper. They designed the control input Q(t) to make the regional average pressure and seismicity-rate outputs track prescribed references. These outputs are defined as regional averages of pressure, yui(t), and seismicity rate, yRi(t).

Dev: So they’re not just looking at one thing; they’re regulating two things at once. Tracking those specific regional averages means the controller has to manage the interaction between how fast the pressure is changing and how fast that change translates into seismic events.

Taro: And I see that this regulation happens under several assumptions, including bounds on the actuation of the plant and constraints on those reference trajectories. That’s a necessary starting point for any control design in a physical system.

Rosa: Yes, and they set up the error dynamics in matrix form as sigma dot t equals psi plus b times q of t, which is how you usually formulate these kinds of control problems mathematically. The controller itself is defined using K1 and K2 terms based on the error state sigma(t) and a term ν(t).

Dev: The controller structure itself is what they call a Multi-Input Multi-Output MIMO Super-Twisting Algorithm, which ensures robustness to uncertainties while generating this continuous control signal. It’s that specific design you’re seeing in their equations, right?

Taro: And I find the Super-Twisting algorithm is interesting because it generates a continuous control signal even if there's a possible discontinuous integral term when l equals negative one. That sounds like they've thought about some of the tricky non-smooth aspects of this system.

Rosa: It shows they’ve accounted for those non-smooth dynamics, which is important because real physical systems aren't always perfectly smooth in their behavior. The results show that the control successfully drives all of the pressure outputs to their smooth references and maintains seismicity rates close to the reference R star.

The paper's improvements: Dev: Moving on, they suggest a few improvements or theoretical guarantees for this approach. They establish closed-loop properties by using explicit bounds on the solution and its time derivative for both the infinite-dimensional dynamics and the nonlinear ODE system. That’s how they prove stability.

Taro: That’s where I want to push a bit: they prove these closed-loop properties yield finite-time or exponential convergence of tracking errors, depending on whether you look at the infinite-dimensional dynamics or the nonlinear ODE system. What does that mean for real control?

Rosa: It means there are theoretical guarantees that the system will converge quickly to those targets, either in finite time or exponentially as time goes on. For the infinite-dimensional dynamics, they prove finite-time stability for l equals negative one.

Dev: And for l between negative one and zero, they show finite-time input-to-state stability with respect to psi bar t. That’s a specific convergence guarantee related to the uncertainty in the system.

Taro: For l equals zero, they prove exponential input-to-state stability with respect to psi bar t. Exponential convergence is generally more desirable for predictable systems when you're trying to manage things in real time, right?

Rosa: Exponential convergence means it’s predictable in terms of how fast the error shrinks once the system reaches a certain state. They also prove that the set of points where the solution rate is unbounded must have zero measure, which is a mathematical way of saying there aren't any wild spikes in behavior they need to worry about.

Conclusion: Rosa: So to wrap up on this paper, this study proposes a mathematical control theoretic approach to mitigating induced seismicity that we haven't seen before. The main implication is that it can significantly reduce the seven hundred twelve events for the Groningen gas-field case without any active control.

Dev: And they also show that parallel CO2 injection can support a more environmentally neutral operation in Scenario two which is combining gas production with CO2 injection toward net-zero carbon goals <ref:2412.06327#pg2>. That’s a practical application that adds another layer to the solution.

Taro: From an autonomy view, this control method gives us a mathematical framework for how autonomous systems can actively prevent induced seismicity by adjusting well fluxes in response to deviations from target outputs. It shows how gracefully the system recovers when things misbehave.

Rosa: I think this work provides a very concrete mathematical control theoretic approach to this challenging problem for the first time, and it suggests that we can use similar PDE-ODE dynamics for various other challenging applications moving forward.

Dev: It’s a solid piece of theoretical work that establishes stability guarantees under uncertainty, which is valuable when you’re designing real-time loops where safety is paramount. We'll keep an eye on how this translates into actual deployment in the field.

Taro: I just think having this kind of mathematical tool for handling the coupling between time and space dynamics is a really useful piece for future research in complex autonomous systems.

Rosa: And that’s our take on "Control of Human-Induced Seismicity in Underground Reservoirs Governed by a Nonlinear PDE-ODE System." We'll be looking at what comes next after this one.

More episodes

← Home