Optimize Your Pipeline Operations With Machine Learning
We have developed a machine learning (ML) decision support solution to help you identify and implement small improvements that will deliver previously unattainable outcomes.
We have developed a machine learning (ML) decision support solution to help you identify and implement small improvements that will deliver previously unattainable outcomes.
Book a Call to Learn How
The nature of pipelines lends itself particularly well to the use of machine learning decision support. Operators must consider a multitude of ever-changing operational variables when managing flow rates. Expecting even senior operators to control the infrastructure for optimal performance in all conditions is not realistic. Decision support gives the operator the recommendations necessary for optimized flow management.
Our client transports crude oil and wanted to test our machine learning decision support solution on a pipeline with six pump stations. Specifically, they wanted to optimize the quantity of drag reducing agent (DRA) used and to decrease electricity usage.
The electricity and DRA costs of operating their single pipeline were reduced by 28.5%. This roughly translates into savings over one million dollars a year for this pipeline.
It is important from the beginning to establish key areas you wish to optimize.
Common Goals Include:
We receive 12 months of data, label the data in the training phase, build a machine learning model, and test the model’s ability to make predictions. For this client, the value assessment estimated 20-40% of cost savings for the pipeline.
We believe in transparency and no empty promises. We share an honest estimate of our machine learning solution’s potential impact. The value assessment must be reviewed and accepted by the client before we move forward with deployment.
How we deploy our ML model is dependent on your risk tolerance. There are three key ways to deploy a machine learning model ranging from decision support to full autonomous optimization:
We are your optimization partners, so we will monitor your infrastructure’s performance as new data accumulates. This allows us to assess the model’s prediction quality and suggest model retraining when it is appropriate. Most ML models require at least one retraining each year that we include with our software license.
If you’d like to learn more about our ML process, please reach out to sales@willowglensystems.com.
The number one reason why artificial intelligence projects fail is lack of expertise (source: MIT Sloan Management Review/Boston Consulting Group survey). It’s only by bringing all three areas of expertise together, that an ML project will be successful.
We leverage our 50-year industrial SCADA expertise when developing our machine learning models. We understand domain specific data and incorporate our knowledge of physics, engineering, and fluid dynamics into our models.
We build and deploy flow computers. So we understand physics, fluid dynamics, and the vast amount of complex data coming from your operations and the pipeline data historian. Our domain expertise grants us the ability to know when to trust the data and how to successfully analyze it.
Labelling and training data sets is hard. Most companies need outside help of some kind. Due to our ML expertise and domain expertise we can accurately and efficiently label the data during the training phase – which is critical in creating supervised machine learning models.
We have hundreds of successful enterprise-level deployments in our history. We are experts in redundancy, fall back, and we follow best practices in deployment. Our enterprise deployment experience provides you peace of mind knowing that your ML project will be rolled out in a controlled and safe manner.