Universiẗat, Diss., 2000--Bremen.

Stuttgart, Univ., Diplomarb., 2001.

In terms of a considerable increase the quality of characterization of tight-gas reservoirs, the aim of the present thesis was (i) an accurate representation of specific conditions in a reservoir simulation model, induced after the hydraulic fracturing or as a result of the underbalanced drilling procedure and (ii) performing the history match on a basis of real field data to calibrate the generated model by identifying the main model parameters and to investigate the different physical mechanisms, e.g. multiphase flow phenomena, affecting the well production performance. Due to the complexity of hydrocarbon reservoirs and the simplified nature of the numerical model, the study of the inverse problems in the stochastic framework provides capabilities using diagnostic statistics to quantify a quality of calibration and reliability of parameter estimates. As shown in the present thesis the statistical criteria for model selection may help the modelers to determine an appropriate level of parameterization and one would like to have as good an approximation of structure of the system as the information permits.

A three-dimensional inversion was implemented for magnetotellurics, which is a passive electromagnetic method in geophysics. It exploits natural electromagnetic fields of the Earth, which function as sources. Their interaction with the conductive parts of the subsurface are registered when components of the electric and the magnetic field are measured and evaluated. The all-at-once approach is an inversion scheme that is relatively new to geophysics. In this approach, the objective function – the basis of each inversion – is called the Lagrangian. It consists of three parts: (i) the data residual norm, (ii) the regularisation part, and (iii) the forward problem. The latter is the significant difference to conventional inversion approaches that are built up of a forward calculation part and an inversion part. In the case of all-at-once, the forward problem is incorporated in the objective function and is therefore already taken into account in each inversion iteration. Thus, an explicit forward calculation is obsolete. As an objective function, the Lagrangian shall reach a minimum and therefore its first and second derivatives are evaluated. Hence, the gradient of the Lagrangian and its Hessian are constituent parts of the KKT system – the Newton-type system that is set up in the all-at-once inversion. Conventional inversion approaches avoid the Hessian because it is a large, dense, not positive definite matrix that is challenging to handle. However, it provides additional information to the inversion, which raises hope for a high quality inversion result. As a first step, the inversion was programmed for the more straightforward one-dimensional magnetotelluric case. This was particularly suitable to become familiar with sQMR – a Krylov subspace method which is essential for the three-dimensional case to be able to work with the Hessian and the resulting KKT system. After the implementation and validation of the one-dimensional forward operator, the Lagrangian and its derivatives were set up to complete the inversion, which successfully solved the KKT system. Accordingly, the three-dimensional forward operator also needed to be implemented and validated, which was done using published data from the 3D-2 COMMEMI model. To realise the inversion, the Lagrangian was assembled and its first and second derivatives were validated with a test that exploits the Taylor expansion. Then, the inversion was initially programmed for the Gauss-Newton approximation where second order information is neglected. Since the system matrix of the Gauss-Newton approximation is positive definite, the solution of this system of equations could be carried out by the conventional solver pcg. Based on that, the complete KKT system (Newton\\\'s method) was set up and preconditioned sQMR solved this system of equations.

In terms of a considerable increase the quality of characterization of tight-gas reservoirs, the aim of the present thesis was (i) an accurate representation of specific conditions in a reservoir simulation model, induced after the hydraulic fracturing or as a result of the underbalanced drilling procedure and (ii) performing the history match on a basis of real field data to calibrate the generated model by identifying the main model parameters and to investigate the different physical mechanisms, e.g. multiphase flow phenomena, affecting the well production performance. Due to the complexity of hydrocarbon reservoirs and the simplified nature of the numerical model, the study of the inverse problems in the stochastic framework provides capabilities using diagnostic statistics to quantify a quality of calibration and reliability of parameter estimates. As shown in the present thesis the statistical criteria for model selection may help the modelers to determine an appropriate level of parameterization and one would like to have as good an approximation of structure of the system as the information permits.

A three-dimensional inversion was implemented for magnetotellurics, which is a passive electromagnetic method in geophysics. It exploits natural electromagnetic fields of the Earth, which function as sources. Their interaction with the conductive parts of the subsurface are registered when components of the electric and the magnetic field are measured and evaluated. The all-at-once approach is an inversion scheme that is relatively new to geophysics. In this approach, the objective function – the basis of each inversion – is called the Lagrangian. It consists of three parts: (i) the data residual norm, (ii) the regularisation part, and (iii) the forward problem. The latter is the significant difference to conventional inversion approaches that are built up of a forward calculation part and an inversion part. In the case of all-at-once, the forward problem is incorporated in the objective function and is therefore already taken into account in each inversion iteration. Thus, an explicit forward calculation is obsolete. As an objective function, the Lagrangian shall reach a minimum and therefore its first and second derivatives are evaluated. Hence, the gradient of the Lagrangian and its Hessian are constituent parts of the KKT system – the Newton-type system that is set up in the all-at-once inversion. Conventional inversion approaches avoid the Hessian because it is a large, dense, not positive definite matrix that is challenging to handle. However, it provides additional information to the inversion, which raises hope for a high quality inversion result. As a first step, the inversion was programmed for the more straightforward one-dimensional magnetotelluric case. This was particularly suitable to become familiar with sQMR – a Krylov subspace method which is essential for the three-dimensional case to be able to work with the Hessian and the resulting KKT system. After the implementation and validation of the one-dimensional forward operator, the Lagrangian and its derivatives were set up to complete the inversion, which successfully solved the KKT system. Accordingly, the three-dimensional forward operator also needed to be implemented and validated, which was done using published data from the 3D-2 COMMEMI model. To realise the inversion, the Lagrangian was assembled and its first and second derivatives were validated with a test that exploits the Taylor expansion. Then, the inversion was initially programmed for the Gauss-Newton approximation where second order information is neglected. Since the system matrix of the Gauss-Newton approximation is positive definite, the solution of this system of equations could be carried out by the conventional solver pcg. Based on that, the complete KKT system (Newton\\\'s method) was set up and preconditioned sQMR solved this system of equations.

The objective of our studies is the combination of electromagnetic and direct current (DC) resistivity methods in a joint inversion approach to improve the reconstruction of a given conductivity distribution. We utilize the distinct sensitivity patterns of different methods to enhance the overall resolution power and ensure a more reliable imaging result. In order to simplify the work with more than one electromagnetic method and establish a flexible and state-of-the-art software basis, we developed new DC resistivity and electromagnetic forward modeling and inversion codes based on finite elements of second order on unstructured grids. The forward operators are verified using analytical solutions and convergence studies before we apply a regularized Gauss-Newton scheme and successfully invert synthetic data sets. Finally, we link both codes with each other in a joint inversion. In contrast to most widely used joint inversion strategies, where different data sets are combined in a single least-squares problem resulting in a large system of equations, we introduce a sequential approach that cycles through the different methods iteratively. This way, we avoid several difficulties such as the determination of the full set of regularization parameters or a weighting of the distinct data sets. The sequential approach makes use of a smoothness regularization operator which penalizes the deviation of the model parameters from a given reference model. In our sequential strategy, we use the result of the preceding individual inversion scheme as reference model for the following one. We successfully apply this approach to synthetic data sets and show that the combination of at least two methods yields a significantly improved parameter model compared to the individual inversion results. / Ziel der vorliegenden Arbeit ist die gemeinsame Inversion (\"joint inversion\") elektromagnetischer und geoelektrischer Daten zur Verbesserung des rekonstruierten Leitfähigkeitsmodells. Dabei nutzen wir die verschiedenartigen Sensitivitäten der Methoden aus, um die Auflösung zu erhöhen und ein zuverlässigeres Ergebnis zu erhalten. Um die Arbeit mit mehr als einer Methode zu vereinfachen und eine flexible Softwarebasis auf dem neuesten Stand der Forschung zu etablieren, wurden zwei Codes zur Modellierung und Inversion geoelektrischer als auch elektromagnetischer Daten neu entwickelt, die mit finiten Elementen zweiter Ordnung auf unstrukturierten Gittern arbeiten. Die Vorwärtsoperatoren werden mithilfe analytischer Lösungen und Konvergenzstudien verifiziert, bevor wir ein regularisiertes Gauß-Newton-Verfahren zur Inversion synthetischer Datensätze anwenden. Im Gegensatz zur meistgenutzten \"joint inversion\"-Strategie, bei der verschiedene Daten in einem einzigen Minimierungsproblem kombiniert werden, was in einem großen Gleichungssystem resultiert, stellen wir schließlich einen sequentiellen Ansatz vor, der zyklisch durch die einzelnen Methoden iteriert. So vermeiden wir u.a. eine komplizierte Wichtung der verschiedenen Daten und die Bestimmung aller Regularisierungsparameter in einem Schritt. Der sequentielle Ansatz wird über die Anwendung einer Glättungsregularisierung umgesetzt, bei der die Abweichung der Modellparameter zu einem gegebenen Referenzmodell bestraft wird. Wir nutzen das Ergebnis der vorangegangenen Einzelinversion als Referenzmodell für die folgende Inversion. Der Ansatz wird erfolgreich auf synthetische Datensätze angewendet und wir zeigen, dass die Kombination von mehreren Methoden eine erhebliche Verbesserung des Inversionsergebnisses im Vergleich zu den Einzelinversionen liefert.

In an effort to improve the accuracy of common 1-D analysis for frequency domain helicopter-borne electromagnetic data at reasonable computing costs, a 3-D inversion approach is developed. The strategy is based on the prior localization of an entire helicopter-borne electromagnetic survey to parts which are actually affected by expected local 3-D anomalies and a separate inversion of those sections of the surveys (cut-&-paste strategy). The discrete forward problem, adapted from the complete Helmholtz equation, is formulated in terms of the secondary electric field employing the finite difference method. The analytical primary field calculation incorporates an interpolation strategy that allows to effectively handle the enormous number of transmitters. For solving the inverse problem, a straightforward Gauss-Newton method and a Tikhonov-type regularization scheme are applied. In addition, different strategies for the restriction of the domain where the inverse problem is solved are used as an implicit regularization. The derived linear least squares problem is solved with Krylov-subspace methods, such as the LSQR algorithm, that are able to deal with the inherent ill-conditioning. As the helicopter-borne electromagnetic problem is characterized by a unique transmitter-receiver relation, an explicit representation of the Jacobian matrix is used. It is shown that this ansatz is the crucial component of the 3-D HEM inversion. Furthermore, a tensor-based formulation is introduced that provides a fast update of the linear system of the forward problem and an effective handling of the sensitivity related algebraic quantities. Based on a synthetic data set of a predefined model problem, different application examples are used to demonstrate the principal functionality of the presented algorithm. Finally, the algorithm is applied to a data set obtained from a real field survey in the Northern German Lowlands. / Die vorliegende Arbeit beschäftigt sich mit der 3-D Inversion von Hubschrauberelektromagnetikdaten im Frequenzbereich. Das vorgestellte Verfahren basiert auf einer vorhergehenden Eingrenzung des Messgebiets auf diejenigen Bereiche, in denen tatsächliche 3-D Strukturen im Untergrund vermutet werden. Die Resultate der 3-D Inversion dieser Teilbereiche können im Anschluss wieder in die Ergebnisse der Auswertung des komplementären Gesamtdatensatzes integriert werden, welche auf herkömmlichen 1-D Verfahren beruht (sog. Cut-&-Paste-Strategie). Die Diskretisierung des Vorwärtsproblems, abgeleitet von einer Sekundärfeldformulierung der vollständigen Helmholtzgleichung, erfolgt mithilfe der Methode der Finiten Differenzen. Zur analytischen Berechnung der zugehörigen Primärfelder wird ein Interpolationsansatz verwendet, welcher den Umgang mit der enorm hohen Anzahl an Quellen ermöglicht. Die Lösung des inversen Problems basiert auf dem Gauß-Newton-Verfahren und dem Tichonow-Regularisierungsansatz. Als Mittel der zusätzlichen impliziten Regularisierung dient eine räumliche Eingrenzung des Gebiets, auf welchem das inverse Problem gelöst wird. Zur iterativen Lösung des zugrundeliegenden Kleinste-Quadrate-Problems werden Krylov-Unterraum-Verfahren, wie der LSQR Algorithmus, verwendet. Aufgrund der charakteristischen Sender-Empfänger-Beziehung wird eine explizit berechnete Jakobimatrix genutzt. Ferner wird eine tensorbasierte Problemformulierung vorgestellt, welche die schnelle Assemblierung leitfähigkeitsabhängiger Systemmatrizen und die effektive Handhabung der zur Berechnung der Jakobimatrix notwendigen algebraischen Größen ermöglicht. Die Funktionalität des beschriebenen Ansatzes wird anhand eines synthetischen Datensatzes zu einem definierten Testproblem überprüft. Abschließend werden Inversionsergebnisse zu Felddaten gezeigt, welche im Norddeutschen Tiefland erhoben worden.

Eine Standardformulierung der Randwertaufgabe für die Beschreibung zeitharmonischer elektromagnetischer Phänomene hat die Vektor-Helmholtzgleichung für das elektrische Feld zur Grundlage. Bei niedrigen Frequenzen führt der große Nullraum des Rotationsoperators zu einem instabilen Lösungsverhalten. Wird die Randwertaufgabe zum Beispiel mit Hilfe der Methode der Finiten Elemente in ein lineares Gleichungssystem überführt, äußert sich die Instabilität in einer schlechten Konditionszahl ihrer Koeffizientenmatrix. Eine stabilere Formulierung wird durch die explizite Berücksichtigung der Kontinuitätsgleichung erreicht. Zur numerischen Lösung der Randwertaufgaben wurde eine Finite-Elemente-Software erstellt. Sie berücksichtigt unter anderem unstrukturierte Gitter, räumlich variable, anisotrope Materialparameter sowie die Erweiterung der Maxwell-Gleichungen durch Perfectly Matched Layers. Die Software wurde anhand von Anwendungen in der marinen Geophysik erfolgreich getestet. Insbesondere demonstriert die Einbeziehung von Seebodentopographie in Form einer stetigen Oberflächentriangulierung die geometrische Flexibilität der Software. / The physics of time-harmonic electromagnetic phenomena can be mathematically described by boundary value problems. A standard approach is based on the vector Helmholtz equation in terms of the electric field. The curl operator involved has a large, non-trivial kernel which leads to an instable solution behaviour at low frequencies. If the boundary value problem is solved approximately using, e. g., the finite element method, the instability expresses itself by a badly conditioned coefficient matrix of the ensuing system of linear equations. A stable formulation is obtained by taking the continuity equation explicitly into account. In order to solve the boundary value problem numerically a finite element software package has been implemented. Its features comprise, amongst others, the treatment of unstructured meshes and piecewise polynomial, anisotropic constitutive parameters as well as the extension of Maxwell’s equations to the Perfectly Matched Layer. Successful application of the software is demonstrated with examples from marine geophysics. In particular, the incorporation of seafloor topography by a continuous surface triangulation illustrates the geometric flexibility of the software.

Eine Standardformulierung der Randwertaufgabe für die Beschreibung zeitharmonischer elektromagnetischer Phänomene hat die Vektor-Helmholtzgleichung für das elektrische Feld zur Grundlage. Bei niedrigen Frequenzen führt der große Nullraum des Rotationsoperators zu einem instabilen Lösungsverhalten. Wird die Randwertaufgabe zum Beispiel mit Hilfe der Methode der Finiten Elemente in ein lineares Gleichungssystem überführt, äußert sich die Instabilität in einer schlechten Konditionszahl ihrer Koeffizientenmatrix. Eine stabilere Formulierung wird durch die explizite Berücksichtigung der Kontinuitätsgleichung erreicht. Zur numerischen Lösung der Randwertaufgaben wurde eine Finite-Elemente-Software erstellt. Sie berücksichtigt unter anderem unstrukturierte Gitter, räumlich variable, anisotrope Materialparameter sowie die Erweiterung der Maxwell-Gleichungen durch Perfectly Matched Layers. Die Software wurde anhand von Anwendungen in der marinen Geophysik erfolgreich getestet. Insbesondere demonstriert die Einbeziehung von Seebodentopographie in Form einer stetigen Oberflächentriangulierung die geometrische Flexibilität der Software. / The physics of time-harmonic electromagnetic phenomena can be mathematically described by boundary value problems. A standard approach is based on the vector Helmholtz equation in terms of the electric field. The curl operator involved has a large, non-trivial kernel which leads to an instable solution behaviour at low frequencies. If the boundary value problem is solved approximately using, e. g., the finite element method, the instability expresses itself by a badly conditioned coefficient matrix of the ensuing system of linear equations. A stable formulation is obtained by taking the continuity equation explicitly into account. In order to solve the boundary value problem numerically a finite element software package has been implemented. Its features comprise, amongst others, the treatment of unstructured meshes and piecewise polynomial, anisotropic constitutive parameters as well as the extension of Maxwell’s equations to the Perfectly Matched Layer. Successful application of the software is demonstrated with examples from marine geophysics. In particular, the incorporation of seafloor topography by a continuous surface triangulation illustrates the geometric flexibility of the software.