AI applications
Partnership
We are collaborating with key technology partners to implement the following AI and Machine Learning initiatives, while we lead the data management efforts to ensure robust and efficient integration.
Smart Energy Engineering
We enable our clients to achieve financial success from oil and gas assets, using analytics and digital transformation.
applications
Key Technology
Well Virtual Metering and Forecasting
This AI application optimizes well performance through data-driven virtual metering and short-term forecasting. It employs Machine Learning models (e.g., SVR, NN, ARIMAX) using real-time sensor data and well tests to predict flow rates and derive optimal operating conditions. It provides automated workflows, achieves over 98% metering accuracy, and delivers a 2.5% net profit gain.
Automated Subsurface Retrieval Engine
An AI-enabled conversational search engine that automates subsurface knowledge retrieval from diverse internal repositories, including unstructured text. It employs text analytics and Natural Language Processing (NLP) using open-source tools like Scikit-learn and Tensorflow. This significantly speeds up information recall, enhances knowledge sharing, and boasts 90% search accuracy.
Data-Driven Production/Injection Optimization
This AI application optimizes production/injection rates in waterflooded fields, maximizing hydrocarbon recovery by balancing Voidage Replacement Ratio (VRR). It uses the Capacitance Resistance Model (CRM) for injection allocation, feeding an in-house optimization procedure. This fast, robust solution decreases water production and increases recovery by maintaining optimal VRR, outperforming conventional simulation.
Well Production Surveillance and Optimization
This AI Advisory System automates production surveillance and optimization. It features a surveillance-by-exception engine using real-time data, soft sensors (ML/proxy models), and an expert system (Bayesian Networks) for automated diagnosis. This generates actionable tickets, providing enhanced situational awareness and unlocking an estimated 2-5% production potential by revealing "invisible losses".
Automated Workover Identification and Ranking
This AI application automates reservoir management workflows for identifying and ranking workover, sidetrack, and infill-drilling candidates. It integrates analytics and machine learning (Bayesian Networks) with physics-based approaches. The system identified 20 times more opportunities in 75% less time, enhancing technical/economic value by 34% with 17% less expenses compared to manual methods.
Prediction of Well Events and Anomalies
This AI application detects abnormal production events in wells, vital when labeled datasets are limited. It leverages Autoencoders (AE), an unsupervised machine learning model, trained on normal behavior to identify anomalies via reconstruction error. Achieving 94% accuracy with 70% precision/recall, it effectively notifies engineers of deviations from expected well performance, even for rare events.