Training

Surveillance, Optimization, Data Analytics and Machine Learning for ESP Engineers

• 1 day / 8 hours
• 2 days or four half-day virtual sessions (extended option)
• Level: Intermediate

Synopsis

Business impact: this course/workshop aims to give an understanding of surveillance, data analytics and machine learning principles as they relate to ESP wells through practical applications. Field data is employed to clarify workflows related to surveillance, optimization and data analysis. Participants assess and extract value from the data sets with the help of easy-to-follow solution scripts. The practical, hands-on approach enables them to confidently test these techniques on their ESP well data.

To ensure profitable ESP well operation with minimal NPT and failure rates, robust surveillance and optimization are crucial. Real-time data streams from heavily instrumented ESP wells can be challenging to manage with traditional methods. The course bridges gaps by using data analytics and machine learning approaches facilitated by generative AI. In this hands-on course, participants learn about surveillance, optimization, data analysis and data science techniques and workflows applied to ESP wells while reviewing code and practicing. The focus is on developing data-driven models while keeping close to the underlying oil and gas production principles.

The following use cases are discussed, covering their business impact, code walkthroughs and solutions:

• ESP state identification
• Virtual flow meter
• Choke flow rate estimation for high-volume wells using an offshore dataset
• Reservoir productivity assessment for unconventional wells
• ESP failure analysis and prediction

Learning outcomes

• After completing the course, participants will have a set of tools and some pathways to model and analyze their ESP wells’ data in the cloud, find trends, and develop data-driven models.

Who should attend

This Intermediate level course is primarily intended for artificial lift, production and facilities engineers and students to enhance their knowledge base, increase technology awareness, and improve facility with different data analysis techniques applied to large data sets.

Prerequisites

• Understanding of petroleum production concepts and specifically ESP design and operations
• Knowledge of Python is not a must but is preferred to get the full benefit
• Trainees need to bring a computer with a Google Chrome or Edge browser and a Google email account (available for free). The Google Colaboratory environment, available in Google Cloud, is used for hands-on exercises.

Course outline

• Digital Oilfield
  • Digital transformation and oilfields
  • Key technologies for digital oilfields
  • Oilfield system data verification and management
• Digitalization in ESP and production optimization
  • ESP surveillance approaches
  • Optimization approaches
  • AI/ML-infused ESP management
• Data types in the production domain: streaming (real-time or time-series) vs. static (non-streaming)
  • Data processing challenges
  • Data basics: cleaning, filtration and regulation
  • Best practices on data exploratory analysis
• AI, ML and Deep Learning: brief and incomplete primer
  • Data analytics lifecycle
  • Bias-variance-complexity tradeoff
  • Data preparation
  • Model types
  • Role of domain knowledge
  • Training and evaluating model
  • Toolsets
• System setup and checks
  • Google CoLab: why do we need it?
  • Pull datasets and codebase from the GitHub repository
• Case studies
  • Problem statement
  • Data requirement
  • Using generative AI in Colab
  • Case study 1: ESP state identification with downhole data / gradient curve predictions
  • Case study 2: VFM using catalog curves (PDP pip)
  • Case study 3: VFM with pressure drop across tubing (PDP to WHPf)
  • Case study 4: reservoir productivity assessment for unconventional well
  • Case study 5: ESP failure analysis approaches

Formats and customization

• Level: Intermediate. One day (8 hours), in person or virtual.
• Training method: the course can be taught virtually, though the best experience and results are achieved through mutual interaction in a classroom setting, which is also conducive to workshop-type outcomes due to increased interaction. For each use case, the instructor shows the solution using a data analysis technique, with the Python code deployed in Google Cloud. Trainees solve a problem and tweak their solution.
• The training can be extended to 2 days or four half-day virtual sessions. The extended option includes five additional use cases.
• Client dataset-based examples are optionally incorporated in the class discussions. This option requires discussions with the client about the problem, two days of consulting effort, and access to the client dataset at least 4 weeks before the class.
• Software: Python code in Google Colaboratory (Google Cloud), with generative AI in Colab; datasets and code pulled from a GitHub repository.

Request this course in-house