by Drs. M.A.C. Kemper, Ikon Science
For SEG Workshop 11: 4D Part 2 - Reservoir surveillance in a "lower for longer" world: Getting more for less – Friday 21st October 2016

ABSTRACT

Reservoir monitoring has seen wide and increasing adoption across the oil and gas community. The objectives of reservoir monitoring are numerous, but from a high level aim to; optimize field development planning, profitably prolonging field life, avoid drilling poor producers, assess remaining hydrocarbon through rock physics techniques and better understand and optimise production forecasting through improved history matching and reservoir model updates.

Commonly, reservoir monitoring takes the form of repeat seismic (and perhaps well log) acquisition and subsequent 4D seismic processing in order to deliver 3D images or snapshots of the subsurface at different stages in the production life cycle. From these snapshots, information about the dynamic behaviour of the reservoir is sought by differencing of the baseline and monitor surveys and extensive interpretation efforts by dedicated teams of geophysical and engineering specialists.

The process of 4D seismic interpretation is a complicated one, requiring a rigorous understanding of engineering principles and processes, their impact on the bulk properties and environmental conditions (stress, temperature) of the reservoir and seal rocks (including the shallow overburden) and consequently their seismic expression(s). In order to make quantitative geophysical and engineering assessments of the impact of a particular development strategy one needs to be able to bridge the gap between often subtle geophysical changes in amplitudes and timing differences and their relationship to underlying geology (typically static) and hydrodynamic and stress (dynamic) states.

This process can typically be achieved through a process of forward modelling (based on engineering data and rock physics techniques) and modern pre-stack seismic inversion technology. To estimates absolute (quantitative) properties from seismic data using conventional seismic inversion techniques, a robust estimation of an initial geological and hydrodynamic model (aka low frequency model) is usually required. This presents a number of particularly challenging problems such as;
  1. How to construct a static geological model which captures reservoir heterogeneity and which also honours (through rock physics transforms and seismic modelling) baseline seismic
  2. How to construct a dynamic model which represents the fluid, pressure and stress distributions within the geological framework at later stages in the production life cycle which also honours (through rock physics transforms and seismic modelling) baseline and monitor seismic.

For the reasons above, many companies favour the use of ‘relative’ 4D inversion workflows – that is inversion without an absolute (low frequency) component. This approach is highly interpretive and prone to significant error and ambiguity.

In this paper we start by presenting a new technique for estimating absolute elastic properties and corresponding facies (litho-fluid rock types) without the pre-requisite low frequency starting model used in other conventional techniques. The model is estimated through the process of iterative seismic inversion, utilising prior rock type dependent rock physics depth trends.

We subsequently show a case study where the new inversion algorithm was applied to multiple 4D seismic surveys covering a giant North Sea oil field to provide independent 3D ‘snapshots’ of the reservoir architecture and fluid distribution through time. The inverted fluid distributions were interpreted and validated by considerable production data made available by the operator and shown to provide robust estimates of reservoir architecture and production effects associated with injector and producer locations over a 25+ year period. We believe this new technique can ‘get more for less’ from legacy 4D datasets.

Finally we have a look ahead. The case study was one where the new inversion was applied twice on two seismic data sets. What would it take to update this new technology to provide a 4D inversion applied on both seismic data sets in one go? What is the potential gain?