Research Abstract

Integrated Asset Management in Smart Oilfields

The delivery of enriched information from technical analysis into real time operational domains is one of the main challenges addressed by Integrated Asset Management (IAM). An IAM system should ensure proper coordination between data collection sources and data processing destinations. Ultimately, meeting these conditions increases the demand for rapid delivery of relevant data to applications at the desired frequency and/or density, and synchronized in time over multiple sources. Large volumes of data from multiple sources result from progressively improving new capabilities for well measurement, seismic data acquisition, and continuous data collection.

Current areas of research in this project are as follows:

1. Metadata Management Using Semantic Web Technologies

One of the goals of Integrated Asset Management(IAM) is to provide improved information access and integration from a variety of tools for reservoir modeling, simulation, and performance prediction to aid rapid decision making for continuous production optimization. Our key hypothesis is that, metadata, which is data about data, plays a key role in making the data more readily accessible and available in more meaningful ways to the users. We demonstrate this hypothesis through a framework for data integration and information access which uses metadata as the main enabling element. The framework consists of two components: a metadata catalog for storage and retrieval of materialized metadata, and a metadata based virtual data integration framework, for accessing non-materialized data. OWL, a recently proposed W3C semantic web standard, which provides a rich set of primitives for data modelling has been used to represent and store the metadata. We examine the opportunities, challenges, techniques and limitations related to using OWL, for realising such a metadata based framework. The key problems of metadata schema design, acquisition/capture, storage, querying and application are addressed. We believe that our work has a wider applicability in domains that contain large amounts of unmanaged silos of data including the grid/eScience applications as well as in many large enterprises.

2. Provenance Management and Integration

Data provenance is metadata that pertains to the derivation history of data objects. Information about how, when, and by whom a piece of data is created and modified, coupled with knowledge about domain processes, allows scientists and engineers to estimate the accuracy and the currency of data. Provenance information is useful in tasks such as data auditing, estimating data quality, and data integration. Our research is focusing on provenance collection, management and integration in the heterogeneous and distributed environment of oil industry. Some research problems of interest include 1) Collecting and modeling provenance from heterogeneous applications and data sources; 2) Integrating distributed and incomplete provenance information to compose complete provenance graph; 3) Effective management and querying of distributed, semantic provenance repositories for energy informatics workflows.

3. Exploration of Semantic Web technologies in Oil Industry:

This research includes two parts: 1) Semantic modeling for facility engineering data. We aim to design semantic models and techniques to provide a cleaner, more maintainable and extensible solution to facility engineering data management. We have successfully demonstrated our designs of semantic models on real world oil field data management systems. 2) Benchmarking semantic web technologies for oil enterprise. We analyzed the adaptability of semantic web technologies in the oil enterprise from the perspective of storage and querying enterprise scale data.

4. Social Network Analysis under the Scope of Enterprise Collaboration Environments:

We investigate the synergy between the underlying network graph and published content under a complex semantics umbrella. We investigate the synergy between the underlying network graph and published content under a complex semantics umbrella. Our current efforts focus on a higher resolution representation that will enable semantically enriched description of social network users, their interests and their relationships to other users. Based on such multifaceted representation we will define complex query operators and apply them in real life applications, thus enhancing current social networking functionality, while at the same time solving scalability and real-time performance issues.

Previous research areas in this project include:

1. Service-oriented architecture

Service oriented architecture (SOA) is a style of architecting software systems by packaging functionalities as services that can be invoked by any service requester. An SOA typically implies a loose coupling between modules, enables software reuse, and leads to scalable and modular architectures. Web services form an attractive basis for implementing service-oriented architectures for distributed systems. We are exploring SOA for our IAM framework. The service abstraction is expected to provide a uniform way to mask a variety of underlying data sources (real-time production data, historical data, model parameters, reports, etc.) and functionalities (simulators, optimizers, sensors, actuators, etc.). Workflows can be composed by coupling service interfaces in the desired order. The workflow specification can be through a graphical or textual front end and the actual service calls can be generated automatically.

2. Workflows for automation and application integration

Engineers spend much of their time performing repetitivetasks which could be automated. Many of these repeated tasks involve invoking legacy tools, preparing data for them, post-processing data from one tool t o be suitable for another etc. A key goal of IAM is to automate such repetitive tasks. We are exploring the use of workflows and workflow techno logies to address this problem (E.g., Windows WF). We envisage an easy-to-use design environement in which a knowledgeable domain engineer can create workflows and publish them to be used by others. Key research issues in creating such a workflow frame work are ability to support creation of domain workflows and monitor the exection of workflows at different levels of hierarchy, creating wrappe rs for legacy tools, integrating legacy tools in a workflow using adapters etc.

3. Uncertainty modeling and design space exploration:

A key requirement for IAM is support for rapid decision-making. The quality of decision-making is significantly impacted by the mechanisms for representation and management of uncertainty in the IAM framework. The space of possible “designs” is characterized by the uncertainty in information about the oilfield asset as well as the range of operational strategies that can be adopted for production optimization etc. One of the current focus areas for this project is a mechanism for rapid, guided exploration of the decision space by using multi-granularity model proxies that allow successive refinement and evaluation of the strategy space. Our multi-level approach will allow users to model uncertainties and to assess the impact of information uncertainty on the decision space. The first level will enable coarse-grained evaluation of the design space using heuristic-based tools, and the second level will use more fine-grained proxies or low-level simulators. The MILAN project represents our prior experience with design space exploration in a model-based framework. Lessons from hierarchical design space exploration in MILAN will be leveraged for this project.

4. Model-integrated computing:

Model integrated computing (MIC) has been adopted as a fundamental design philosophy for our IAM framework. An MIC-based approach involves rich, domain-specific modeling languages and environments that capture system information in a structured format. This model information (model database) forms the core of the system and acts as the central point of information sharing and coordination between loosely coupled applications. Model information can be utilized for a variety of workflows from automatic code generation, workflow synthesis, etc. Our early work on defining a modeling approach for the petroleum domain is described in this paper. See the Publications section.

5. Domain-specific languages:

An important area of ongoing research is the design of a domain specific modeling language (based on a UML-like notation) for capturing various aspects of IAM. This modeling language should capture not just the information from the petroleum domain (such as reservoir and well information, fluid properties, etc.) but also the computational basis for IAM such as databases, web service interfaces, XML-based data schemas, simulators, optimizers, real-time sense-and-response workflows, etc. The eventual objective is to allow an end user to describe oilfield asset components, compose the components into operational strategies, and invoke a set of integrated tools for specific workflows that support the strategies. The design of the domain-specific language is a perpetual and critical component of the framework design and implementation.