Application of Artificial Neural Network to the Realm of Oil & Gas

DESCRIPTION:

This course introduces the applications of Artificial Neural Network in the oil & gas industry. The basic principles covered can be applied to any oil and gas field or unit data.

Course Level: Foundation

THIS COURSE IS DESIGNED FOR YOU, IF YOU ARE:

• An engineer working in oil & gas fields/units

Prerequisite: Participants having basic knowledge of calculus will derive maximum benefit from the course.

HOW WE BUILD YOUR CONFIDENCE:

• Highly interactive course
• Practical oil & gas case examples will be presented and adequate exposure will be provided to the participants through exercises
• The instructor is experienced in the application of neural networks as well as training engineers on the basics of Artificial Neural Networks

THE BENEFITS FROM ATTENDING:

By the end of the course you will feel confident in your understanding of:

• Basic structures of the Artificial Neural Network model
• Managing data appropriately for the neural network model
• Applications of the ANN technique to unknown input data
• Developing ANN models to predict output data

TOPICS:

• Applications of the neural model
• Similarities to biological neurons
• Concept and structure of Artificial Neural Network
• Managing the data for a neural network
• Linear and non-linear separability
• Commonly used activation functions
• Effect of parameters on the activation functions
• Utilising activation functions for mapping outputs - forward pass for a small data set
• Developing the network with a different set of random weights
• Determining the cost function on a small data set based on activation functions
• Basic concept for calculation of derivates of training dataset
• The local & global Minima - avoiding the local minima using momentum term
• Procedure for backpropagation to develop a new set of weights
• Minimisation of cost function using excel solver
• Checking cost function in the validation data set with activation functions
• Finalising the weights for the model
• Testing the finalised weights on validation & test data using different activation functions
• The concept of Multiple Input Multiple Output (MIMO) and Multiple Input Single Output (MISO) models
• Obtain final weights using Leaky ReLu method
• Testing the finalised weights on validation and test data in MIMO & MISO models
• Calculating the cost function on validation and test data
• Computation of cost function based on the parametric ReLu (as per the specified negative gradient) function on pre-defined set of data
• Trial on the test data with model obtained from training dataset
• Applications of neural network to above system using the Swish function
• Gas Compression example with Swish Function
• Mean squared Error with the weights obtained (using Leaky ReLu and Swish activation functions on training data set) on validation and test model
• Review the graphical correlation between the true and model predicted values