Search

We design a leak detection, size estimation and localization algorithm for a branched pipe system, requiring flow and pressure measurements to be taken at the inlet and outlet boundaries, only. By showing that the pipe system model can be mapped into a system of coupled, linear hyperbolic PDEs (partial differential equations), with the parametric uncertainties caused by the leak appearing in a particular way, established methods can be applied to obtain state and parameter estimates. Analyzing the structure of the parametric uncertainties that appear in the obtained adaptive observer canonical form, we prove that total leak size can be estimated regardless of how leaks are distributed in the pipe network. Moreover, any number of point leaks can be located, provided they occur sufficiently separated in time.

We present a novel streaming learning approach, utilizing a deep neural network (DNN) to learn from data available during operation to estimate at-bit density using drilling parameters. Since every wellbore is different, the relationship between drilling parameters and at-bit density varies. Equipment used, well trajectory, friction and bit wear are examples of conditions that affect this relationship and makes a pre-trained model unable to represent an accurate input/output mapping applicable to all wells. However, using delayed density log measurements, continuously supervising updates to the model is possible during operation. The algorithm has been tested on drilling data from wells on a field operated by Equinor and compared to a standard deep learning approach, where results show that a streaming learning approach outperforms the traditional method. Statistical analyses have been performed to verify the statistical significance and effect size on the data sets. Data visualizations using a t-distributed Stochastic Neighbor Embedding (t-SNE) indicate that the relationship between drilling parameters and density log indeed vary between wellbores, making generalizability an issue for a traditional supervised learning approach to this problem, and motivating a streaming learning approach. Using the proposed method, more accurate at-bit estimates can be made, providing preliminary indications ahead of the tool placed 20–30 m behind the bit, which, dependent on rate of penetration (ROP), will be available 20–120min later.

We consider a scalar 1-D linear hyperbolic partial differential equation (PDE) for which infinite-dimensional backstepping controllers have previously been designed based on boundary actuation and sensing, and incorporate first order actuator and sensor dynamics into the design. Two observer designs are proposed, and combined with a state-feedback into output-feedback control laws which render the origin of the closed-loop system exponentially stable with arbitrary convergence rate. The theory is verified in simulations.

We propose a simple model of two-phase gas–liquid flow by imposing a quasi-equilibrium on the mixture momentum balance of the classical transient drift-flux model. This reduces the model to a single hyperbolic PDE, describing the void wave, coupled with two static relations giving the void wave velocity from the now static momentum balance. Exploiting this, the new model uses a single distributed state, the void fraction, and with a suggested approximation of the two remaining static relations, all closure relations are given explicitly in, or as quadrature of functions of, the void fraction and exogenous variables. This makes model implementation, simulation and analysis very fast, simple and robust. Consequently, the proposed model is well-suited for model-based control and estimation applications concerning two-phase gas–liquid flow. . (c) 2017 This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ available Julu 2018

The paper presents an observer design for a class of hyperbolic PDE–ODE cascade systems with a boundary measurement. The cascade systems consist of coupled PDEs, featuring one rightward and one leftward convecting first-order transport PDEs, and a set of ODEs, which enter the PDEs through the left boundary of the systems. The design, which is based on the Volterra integral transformation, relies only on a single sensor at the right boundary of the system. The observer consists of a copy of the plant plus output injection terms both in the PDEs and the ODEs. The observer is constructed in a collocated setup, which means both sensing and actuation are located at the same boundary. The observer gains are computed analytically by solving Goursat-type PDEs in terms of Bessel function of the first kind. The observer design is tested against a field scale flow-loop test experiment in Stavanger by Statoil Oil Company. The results show that the observer converges to the actual values and that the design can be used as a process monitoring tool in oil well drilling. © 2016. This is the authors’ accepted and refereed manuscript to the article. LOCKED until 22.2.2018 due to copyright restrictions. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/

The ability to perform accurate pore pressure and reservoir inflow estimation during a kick incident is necessary, particularly when drilling in formations with narrow pressure margins. Currently available techniques for pore pressure estimation and reservoir characterization either rely on empirical correlations requiring access to well logging data and other petrophysical information, or require downhole pressure sensing and advanced flow metering capabilities. This paper introduces a model-based estimation technique which uses surface measurements commonly available in a Managed Pressure Drilling (MPD) system, coupled with a simplified transient two-phase model. This model is capable of representing essential dynamics during a gas kick with reduced computational overhead, but without sacrificing significant modeling accuracy. First, the model is validated in a gas kick scenario against experimental data, showing good agreement between key measured parameters and the model predictions, and thereby justifying the model applicability to field operations. Next, data generated from a commercial simulator test case is used to evaluate the proposed estimation methodology. The estimated pore pressure and reservoir productivity are close to their respective values from the commercial simulator, and the flow out rate and surface back-pressure predicted by the simplified two-phase model yield very good match against the simulator results. © 2016 Elsevier B.V. All rights reserved. This is the authors' accepted and refereed manuscript to the article. Locked until 31 May 2018 due to copyright restrictions

As oil exploration and development costs rise, the oil industry increases its efforts to improve oil recovery (IOR) from existing fields. IOR is achieved mainly by drilling more wells, but drilling in partially depleted reservoirs is challenging due to narrow pressure margins. Offshore drilling in harsh environments, such as the North Sea, presents additional challenges, since the heaving motion from a floating rig induces large surge and swab pressures in the well. The paper suggests a remedy for this problem using automatic control of well pressure. Taking advantage of an experimental lab facility recently completed at NTNU, a model of the drilling system is developed using subspace identification methods. The model serves as a basis for state estimation and controller design using model predictive control. Applying the controller to the lab facility, pressure oscillations are suppressed by 70-90% compared to the open-loop case, depending on the period of the heave motion. © 2016. This is the authors’ accepted and refereed manuscript to the article. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/