Oil and Gas – Smart Exploration through Machine Learning and AI
Why Machine learning and AI in Oil and Gas Exploration
O&G drilling has become extremely competitive in recent years
Entry of newer players like Shale operators has impacted the demand supply balance, forcing down market prices
Lowering breakeven points through better prospecting and cost effective operations is imperative to maintain competitiveness
Geological features are extremely complex and often contain non-linear relationships between variables of interest
Similarly, drilling equipment contain several expensive and critical moving parts with complex failure scenarios that need to be predicted realtime
New advanced algorithms like deep neural networks are essential to model such non-linearities
Visualizing Oil Field Surveys
Modern geological surveys capture vast amounts of raw seismic trace data
Deep learning is required to create accurate 3D maps of underground faults and other subsurface features to effectively characterize reservoirs
Without deep learning, the required accuracies are not feasible
Features like natural porosity, permeability, etc. have quite non- linear relations with the outcome of interest, the oil volume in a potential reservoir
Accurate visualization leads to several business advantagese
Effective lease bidding
Higher service revenues for operators
Extracting value from shale formations
Every shale formation is somewhat unique in terms of geological characteristics and requires customized drilling
To improve well productivity and increase ROI, it is important to find effective fracture recipes for each potential site
AI solutions can help find the right mix of spacings, depth of horizontal bores, proppants, and pressure patterns for each well, allowing efficient fracking
AI driven fracking becomes especially important in the mature stage of a well, to extend the lifetime and maintain a more consistent throughput
Maintenance Analytics for Drilling Equipment
Massive amounts of sensor data is collected during drilling such as pump pressures, RPMs, flow rates and subsurface temperatures; yet detecting equipment failure is a difficult problem
Failure scenarios are often chaotic and complex and advanced machine learning is required to generate real time accurate predictions
Predicting equipment health leads to actionability
Preventive Maintenance to extend equipment life
3D view of equipment problems like metal fatigue, rust, corrosion, etc. to allow quicker repair
Temporarily pausing operations to prevent chaotic machine breakdowns and damage to the well
