Reservoir characterisation and monitoring

Time-lapse or 4D seismic is a powerful technology used to optimise hydrocarbon production and monitor CO₂ geological storage. However, the technology is seldom used onshore, in large part due to the presence of time-lapse noise caused by changes of near surface conditions and shallow subsurface properties.

In the framework of CRC for greenhouse gas technologies, CRGC researchers are involved in the first integrated land time-lapse seismic project in Australia. The project involves monitoring of CO₂ sequestration using 4D seismic data over Naylor field, Otway Basin, Victoria. The work involves research in the following topics:

• Analysis of land 2D and 3D seismic repeatability with field tests using surface and buried geophones;
• Developing time-lapse processing flows and specialised algorithms for consistent processing of legacy land data shot with different sources and in different weather conditions;
• Analysis and modelling of time-lapse noise; Rock physics modelling of changes of elastic properties of reservoir rocks from flow simulations; Seismic forward modelling of time-lapse seismic signal from flow simulations;
• Assessment of feasibility of time-lapse monitoring using integration of geological modelling, flow simulations, rock physics, seismic forward modelling, and seismic noise estimation;
• Data acquisition using novel technologies, including buried geophones, permanent sources, and distributed acoustic sensors.

Researchers: Prof. Roman Pevzner, Prof. Milovan Urosevic, Prof. Boris Gurevich

4D surface seismic is a power technology for monitoring and optimisation of hydrocarbon production and CO₂ geological storage. However in some environments, standard 4D seismic is challenging due to relatively high cost, significant time delay between data acquisition and interpretable results, long intervals between surveys, and land access restrictions. To overcome these limitations, CRGC is developing and testing an alternative monitoring strategy involving time-lapse borehole seismic surveys acquired with Distributed Acoustic Sensors installed in multiple wells and permanent seismic sources.

The work involves research in the following topics:

• Analysis of zero offset, offset and multi offset 3D VSP repeatability with field tests using  geophones and fibre-optic cables in various settings;
• Developing processing flows and specialised algorithms for consistent processing of the data for various sources;
• Assessment of feasibility of borehole-based time-lapse monitoring using integration of geological modelling, and surface seismic monitoring.
• Data acquisition using novel technologies, including permanent sources, and distributed acoustic sensors.

Researchers: Prof. Roman Pevzner, Dr. Konstantin Tertyshnikov, Dr. Sinem Yavuz

Onshore seismic surveys typically deploy seismic receiver arrays and mobile sources to image the subsurface. Time-lapse (4D) surveys rely on accurate positioning of source points and receivers in order to monitor changes in the reservoir. Conventional seismic sources with an increased source effort are costly and are likely to leave a noticeable environmental footprint. Furthermore, onshore seismic operations require significant labour, as a large amount of seismic equipment needs to be deployed and then retrieved for each survey. CRGC is exploring the use of permanent seismic receivers and sources in order to overcome limitations of the conventional approach.

To reduce the cost and land impact compared to vibroseis sources, permanent installation of Surface orbital Vibrators (SOVs) can be utilised in monitoring surveys. SOVs consist of common AC induction motors. They produce vibrations as an effect of the rotation of eccentric weights, which produces both a vertical and horizontal shear force. The force of the source increases as frequency squared. With their low production and operating cost and adequate force, SOVs can be a good alternative to common seismic source.  In CO2CRC Otway project, Curtin team optimised permanent SOVs coupled up with conventional geophones as well as Distributed Acoustic Sensors to acquire high quality surface and borehole TL seismic data at relatively low cost and a minimum land impact.

Researchers: Prof. Roman Pevzner, Dr. Sinem Yavuz, Dr. Konstantin Tertyshnikov

Sea-bed electromagnetic (EM) methods are a relatively new technology for hydrocarbon exploration in deep water settings. When utilised in conjunction with seismic data, EM methods have the potential to provide valuable information about the reservoir potential of hosting hydrocarbon.  CRGC research includes survey planning, inversion, and interpretation.

Modern surveying equipment provides the possibility of multi-offset, multi-frequency and more recently multi-azimuth data sets. The arrangement and orientation of EM sources and receivers relative to the geoelectrical nature of the target and host is a key part in determining the success of a Sea Bed Electromagnetic survey. Further analysis of complex Sea Bed EM data sets requires the assistance of inversion. The CRGC research in Sea Bed EM is focused on: (a) integrated visualisation and survey planning; (b) interactive 2D/3D inversion of Sea Bed Electromagnetic data; (c) design of novel instrumentation; and, (d) understanding the impacts of electrical anisotropy on the Sea Bed EM response.

Researchers:  Prof. Brett Harris, Dr. Andrew Pethick

CRGC has developed a machine learning approach to reservoir prediction and adaptive model updates using 4D seismic data which simplifies new data assimilation.

Seismic monitoring of CO₂ sequestration or hydrocarbon production generates data that can be used to continuously update the reservoir model via a seismic history matching process. Standard practice involves iterative runs of reservoir simulations and substantial manual input into model updates. Such a tedious process is impractical for real-time monitoring. To overcome this problem, the Curtin team has developed a machine learning approach to reservoir prediction and adaptive model update. The algorithm is based on the latest developments in machine vision, namely generative convolutional networks and recurrent convolutional networks, and is trained using numerous realisations of standard flow simulations. The algorithm is assessed using flow simulation not included in the training set and this comparison gives promising results.

Researchers: Dr. Stanislav Glubokovskikh, Prof. Roman Pevzner