First Break - Volume 42, Issue 1, 2024

Petrophysical modelling is important in reservoir characterisation. Classic geostatistical approaches have been widely used to generate 3D petrophysical properties. However, many manual interactions are required in classic approaches because they are based on the simple stationarity assumption. Various machine-learning (ML) algorithms have been developed to reduce the cycles of petrophysical modelling.

In this paper, we apply a ML petrophysical modelling algorithm combining random forests and Kriging to the Groningen gas field. Four scenarios are evaluated: (1) using different quality of well data as input, (2) using different geometrical variables as secondary variables, (3) using different seismic attributes as secondary variables, (4) upscaled gamma and density properties with a nonlinear relationship. The conclusions are (1) bad-quality well data can seriously impact the results, (2) inclusion of more geometrical variables can dramatically improve the results, (3) the impact of seismic attributes on the results heavily depends on the correlation between seismic attributes and input wells, (4) generated gamma and density properties can automatically reproduce the nonlinear relationship in wells. This case study is helpful both for better use of this algorithm in future studies and for providing a reference for evaluating other ML petrophysical modelling algorithms based on the Groningen dataset.

The authors describe how in recent onshore nodal seismic data acquisition surveys IoT technology has been used to efficiently provide both cable-free Ambient Noise Monitoring widely sampled across the recording spread, to enable informed decision making regarding noise levels affecting recording operations, and the mitigation of safety risk to remote workers, especially in hazardous terrain, by providing real time status monitoring and an emergency SOS capability where traditional communications are limited or impaired.

Since the inception of the seismic reflection method in the 1920s, equipment availability and limits in technology (not forgetting cost) have restricted the acquisition methods needed to achieve a desired level of ‘quality’ in any survey. Although great changes have occurred over recent decades, the aim of any survey remains the same; to obtain the highest ‘quality’ of data by using the best methods and equipment available to achieve the greatest enhancement of ‘signal’ and/or reduction of the ‘noise’ (signal-to-noise ratio) during acquisition. Recent technological improvements in the sensors, positional and recording equipment, have also permitted greater productivity and a decrease in cost per recording channel – survey designs are now less dependent on equipment limitations than in previous decades. The sort of equipment offered by most current manufacturers means that the primary method of acquiring the ‘best quality’ data is by means of high fold/high density surveys utilising a single source, single sensor and a multitude of ‘blended sources’ to achieve both the required SNR and productivity. Increasing the ‘quality’ of a survey is now synonymous only with increasing fold (number of traces that fall within a designated area). While it is undeniable that high values of fold and high source and receiver densities can be very effective, this ‘brute force approach’ relies on a multiplicity of shots and receivers to improve the SNR with little consideration of the source ‘strength’ (the energy / amplitude of the propagating seismic signal recorded in each trace). In this article, we discuss how the signal strength of a source affects the SNR, fold, and productivity.

The all-digital MEMS (Micro-electromechanical system) based sensor designed for seismic acquisition was first introduced in the 1990s. Early implementations of the sensor showed great promise in ocean bottom cables where properties such as low-frequency sensitivity and tilt insensitivity were particularly advantageous. Further implementation of the technology on land yielded three-component sensors (P and S wave) and finally single-component sensors (P wave) that are currently utilised on land production projects in several regions of the world. The seismic MEMS-based sensor has matured into a time-tested, highly refined tool for seismic exploration. The next evolution is merging of the MEMS sensor with the standard land seismic nodes. Until recently, all seismic nodes have been implemented and deployed with analog sensors. However, nodes are currently getting deployed with MEMS-based sensors combining the benefits of both technologies to form a unified solution for seismic acquisition.