Advanced Data Analytics and Deep Learning for Predicting Formation Tops
A great example of how technological collaboration between Tape Ark and DataVediK can liberate large volumes of previously inaccessible exploration data to the public cloud (in this case well log data) and then through the application of dataVediK’s artificial intelligence and machine learning technologies empower oil and gas and exploration organisations to increase operational efficiencies and reduce overall costs.
The Challenge
Petabytes of data that have not been accessed for years, all kept because of both the cost to acquire and the strong sense that it may be useful again. Every oil company has it, but most don’t know what to do with it.
Historically, oil and gas companies have archived their data off to a tape holding in offsite storage. If they need the data again, they are usually happy to wait for days or weeks to get access to the data that they want. Time and cost usually prevents oil companies from working on masses of data that they don’t consider core to their current needs.
The rapidly changing big data and analytics industry is now showing mind blowing results on massive scales of data that is changing the way oil and gas companies view how their archive should be handled. One of the first things that shows itself to be a critical component to the new thinking is the use of private clouds and private data storage solutions where access to the data is not possible, impractical or too costly are rapidly going out of favour. The use of public cloud is now becoming acceptable to oil and gas companies and data in record amounts are flowing into the public cloud, and out of private clouds all over the world. It is safe to say that anyone putting data into a private cloud is going to need to stop and pivot to keep pace with the technology being developed in the public arena.
The Tape Ark & dataVediK Solution
Tape Ark has set its sights high with a plan to liberate the world’s second largest collection of data – that which is on tape in offsite vaults and in private clouds. Moving this data to the public cloud is seen as essential by the Tape Ark team to allow new technology to put this data to use by data analytics and big data tool sets.
One example of how the liberation of a large well data collection can be achieved by data analytics and machine learning to gain new knowledge has been done by dataVediK. Houston based dataVediK (a name based on the Sanskrit word “Veda” which means knowledge and wisdom), a Tape Ark partner, is an early stage startup specialising in Big Data & Analytics, Machine Learning, Automation and end-to-end Data Ecosystems for Oil & Gas industry. dataVediK aims to decode the knowledge hidden in data by using data analytics and machine learning technologies on oil and gas datasets.
dataVediK’ s mission is to solve Oil & Gas business problems by blending machine learning techniques with E&P domain expertise and delivering solutions embedded in Geoscience, Production and Drilling workflows. dataVediK also combines the Big Data processing techniques with principles of data management & data quality management (DQM) to deliver end-to-end solutions using innovative user-centric visualisation with flexible deployment architectures including public-cloud, on-premises and hybrid environments.
With the enormously experienced team in upstream domain, dataVediK has developed machine learning solutions for diverse problems encountered in the oil and gas industry. dataVediK’s repeatable process helps in building and training machine learning models rapidly and in a consistent fashion on a variety of datasets.
Tangible Results
dataVediK recently demonstrated a workflow to predict formation tops by applying artificial intelligence and machine learning techniques to learn the well logs signatures. This deep learning model provides high quality predictions to aid geologists in picking lithology markers consistently and in an accelerated fashion, thus boosting their overall operational efficiency. The self-learning model, which is a unique differentiator of dataVediK, encompasses the detection of outliers and data quality issues and their subsequent validation and suggested corrections to improve the quality of data in an automated fashion during the model training process. As a result, the model learns from good data only and delivers results with unparalleled accuracy which makes it very appealing to not only the end-user geoscientists, but also to the data management groups and the decision making executive stakeholders.
Example of Object, Activity and Scene Detection
Without the liberation of large volumes of historical well log data from tape into the public cloud by Tape Ark, this kind of data modelling and applied machine learning by dataVediK is simply not feasible on the same scale. By making this archive collection of oil and gas data available and accessible, organisations can apply machine learning solutions and experiment with and develop further technological solutions to achieve reduction of costs, better use of resources, increased operational efficiencies.
