Select Page
Home 9 Training 9 Course Listing 9 Artificial Neural Systems in Petroleum Engineering
Artificial Neural Systems in Petroleum Engineering
John C. Lorenz, PhD
INSTRUCTOR: Robert F. Shelley, PE
DISCIPLINE: Engineering
CEUS: 1.6
AVAILABILITY: Public & In-House 
WHO SHOULD ATTEND: Anyone who is frustrated with the limitations or inconvenience of physics-based models. Predictive Neural Network Models (ANN) developed from data provide quick answers, offer a fresh perspective, test physics-based modeling assumptions and can quantify the impact of geology, chemistry, operations and other factors on well production and recovery.

This course is designed for experienced professionals charged with completion/frac design, completion/frac evaluation, prospect evaluation, recompletion candidate selection, general solution development, identifying and quantifying the significance of parameters, or forecasting outcomes such as sand-outs, well production, recovery, etc. Potential participants include engineers, geologists, geophysicists, managers, data scientists, technicians, chemists, etc.

Participants in this class will obtain the knowledge and expertise needed to build models to make predictions and quantify the impact of cause-and-effect relationships. Programming skills are not required for successful completion of this course.

COURSE DESCRIPTION: Participants will learn the fundamentals and develop a working knowledge of ANN modeling technology through the course’s hands-on model building exercises. An Artificial Neural Network is a form of Artificial Intelligence (AI) that can develop a relationship between known outputs and a desired outcome. ANN modeling case histories will be studied to help participants understand why this approach was used and provide examples in which significant cost savings and value has been created. These examples include predicting well production, recovery, completion design, hydraulic fracturing, reservoir classification and prospect evaluation. Additionally, participants will obtain a working knowledge about building models from data and be able to identify potential applications for ANN modeling technology.

All students will be required to have a laptop (with admin rights) with Windows and Excel installed. As a learning aid, a demo version of ANN model development software will be installed during class. This course does not include a software license for students.

Although this is not a programming course, a proficiency with Excel is recommended.


  • Identify potential applications and for the ANN modeling approach.
  • Basic understanding of ANN modeling concepts.
  • Hands on exposure with ANN model development.
  • Model development strategies to address specific needs or problems.
  • Techniques to mitigate real world uncertainty.
  • Strategies to minimize the impact of data issues such as quantity, quality, completeness and consistency.


Day 1

  • Modeling Background
  • Concept and Theory
  • Data Issues and Preprocessing
  • Predictive Model Development
    • Feed Forward Neural Network
      • Model Training and Fitness
      • Predictor Evaluation, Genetic Algorithms
      • Strategies and Applications
      • Case Histories
    • Self-Organizing Mapping
      • Model Training
      • Strategies and Application
      • Case Histories
    • Modeling Experiments
      • Data Quantity
      • Use of Testing Data
      • Predictor Evaluation and Model Selection

Day 2

  • Model Development Exercises
    • Predicting Well Production, Granite Wash Case
      • Double Well Production for Less Cost
    • Completion Design Evaluation, Marcellus Case
      • Facilitate increasing PUD Recovery to 10 BCF/well
    • Evaluate Reservoir Proxy Data, Bakken Case
      • Use Measurements Taken during Horizontal Drilling Operations as Proxies for Reservoir Quality.
    • Predicting Sand-outs, Mission Canyon Case
      • Predict Frac Treatment Sand-out to improve Operational Efficiency
    • Reservoir Classification, Wolfcamp B Case
      • Evaluate the Usefulness of Reservoir Characteristics