Planning a well for safer and faster operations requires complex and experience-based decision making. In this aspect of the oil and gas industry, the main source of information is the limited amount of offset well data. The offset well data is spread across different domains which makes it more difficult to analyze in the stipulated period of time and in turn the quality of a planned well is always sub-optimum. The main objective of this study is to make use of machine learning and deep learning algorithms to increase the scope of offset well planning by analyzing a large number of wells efficiently. This approach represents an alternative workflow for offset well planning with a focus on data-driven KPI's (Key Performance Indicators). The significant KPI's that can be tracked is increased use of reference wells with a drastic reduction in time required for waiting and searching for data.
The area of this study is the Norwegian Continental Shelf (NCS) encompassing wells in the North Sea, the Norwegian Sea, and the Barents Sea. The dataset consists of around 4500 wells with high dimensionality features based on stratigraphy data and well path data (eg. top and bottom depth, thickness, azimuth, inclination etc.) for each well. A new methodology is discussed in this study that includes the use of deep learning algorithms i.e. Convolutional Neural Networks (CNN) and non-linear algorithm for dimensionality reduction (t-Distributed Stochastic Neighbouring Entities and Uniform Manifold Approximation And Projection) on this high dimensional dataset.
The results of this study show that the reduced dimension data obtained from the above methodology is able to extract high-level features from the wells based on the input data/features. Obtained embeddings show the relative distance between the wells which can then be used to find how similar the wells are with respect to each other.
The value of the results obtained from this study can potentially increase the horizon for offset well planning by incorporating not just the nearby wells but all the wells from the corporate databases. This, in turn, tracks key KPI's and saves time by providing automated answers to the engineers.