How to use machine learning to optimize your production -

How can machine learning be used to optimize production?

Production optimization - How can you use machine learning to optimize your production?

Fully autonomous production plants will be available in the not too distant future. But even today, machine learning can make a big difference in production optimization. Here, I'll take a closer look at a concrete example of how machine learning (ML) and analytics can be used to solve a complex problem encountered in a real-life context.

What is production optimization?

Product optimization is a common problem in many sectors. In our context, optimization is any act, process or methodology that makes something - such as a design, system or decision - as good, functional or efficient as possible.

Decision-making processes for minimum cost; optimum quality, performance and energy consumption are examples of such optimization.

To make this more concrete, I'll focus on a case we worked on with a global oil and gas company. Currently, the industry's main focus is on digitization and analysis. This focus is fueled by the vast amounts of data that are accumulated from thousands of sensors every day; even on a single production facility.

Until recently, the use of these data was limited due to the..:

-Competence limits .

-Lack of technology and data pipelines.

In the context of the oil and gas industry, production optimization is essentially "production control": you minimize, maximize or target the production of oil, gas and perhaps water. Your goal might be to maximize oil production while minimizing water production. Or it might be to run oil production and diesel ratio at specified setpoints to maintain desired reservoir conditions.

How complicated is it to optimize production?

Oil and gas production is a complex process, and many decisions have to be made to achieve short, medium and long-term objectives; from planning and asset management to small corrective actions.

Short-term decisions have to be made in a matter of hours, and are often characterized as day-to-day production optimization. They generally seek to maximize oil and gas rates; by optimizing the various parameters controlling the production process.

In most cases today, the day-to-day optimization of production is carried out by the operators controlling the offshore production facility. This optimization is a very complex task where a large number of controllable parameters all affect production in one way or another. Somewhere in the order of 100 different control parameters have to be adjusted to find the best combination of all variables.

In our example, only two controllable parameters affect your production rate: "variable 1;" and "variable 2;". The optimization problem is to find the optimal combination of these parameters; in order to maximize the production rate. Solving this two-dimensional optimization problem isn't all that complicated; but imagine this problem scaled up to 100 dimensions instead.

Now it's a different story. This is essentially what operators are trying to do when they optimize production. Today, the quality of this operation depends to a large extent on the operators' previous experience and understanding of the process they are controlling.

Machine learning (ML) algorithms can accumulate unlimited data sets

This is where an approach based on machine learning (ML) really comes into its own. Optimization by operators is largely based on their own experience, which accumulates over time as they become more familiar with process plant control.

This ability to learn from previous experience is exactly what is so intriguing about machine learning. By analyzing vast amounts of historical data from rig sensors; algorithms can learn to understand the complex relationships between different parameters and their effect on production.

The fact that algorithms learn from experience is similar in principle to the way operators learn to control the process. However, unlike a human operator; machine learning (ML) algorithms; have no problem analyzing complete historical data sets for hundreds of sensors over a period of years. They can accumulate unlimited experience compared to a human brain.

How a production optimization algorithm works

Having a machine learning algorithm capable of predicting production rate based on the control parameters you adjust is an extremely valuable tool. The prediction model based on machine learning provides us with a production rate curve with its peaks and troughs representing high and low production.

The multidimensional optimization algorithm then moves through this landscape in search of the highest peak representing the highest possible production rate.

By moving through the production rate curve, the algorithm can give recommendations on how best to reach this peak; i.e. which control variables to adjust, and how far to adjust them. Such a machine learning-based production optimization is therefore made up of three main components:

1. Prediction algorithm:

Your first important step is to ensure that you have a machine learning (ML) algorithm capable of successfully predicting the correct production rates given the parameters of all operator-controllable variables.

2. Multidimensional optimization:

You can use the prediction algorithm as the basis for an optimization algorithm, which explores the control variables to be adjusted in order to maximize production.

3. Usable output:

As a result of the optimization algorithm, you get recommendations on which control variables to adjust, and the potential production rate improvement based on these adjustments.

A machine learning-based optimization algorithm can run on real-time data streaming from the production facility; providing recommendations to operators when it identifies potential for production improvement.

This machine-learning-based optimization algorithm can be used as a tool to assist operators controlling the process, helping them to make more informed decisions to maximize production.

Fully autonomous operation of production facilities is still in the future. Until then, assistance tools based on machine learning can have a substantial impact on the way production optimization is carried out.

In the future, I think machine learning will be used in many more ways than we can imagine today. What impact do you think this will have on different industries?

Author: NAJI Faouzi

Need an expert opinion?

Follow our innovations on social networks

We frequently publish on social networks (Linkedin, Twitter and Medium) our innovations and the new functionalities of our industrial management solutions.

Also, we would be happy to share with you the latest trends in industrial management 4.0 through high quality content that you could share with others.