Internationally, Jamaica is well known for both sports, culture and the high cost of electricity; even Solomon Island is worried about their #1 spot.
The high cost of electricity in combination with the other economic factors has resulted in an increase in the number of theft. It has been reported that theft by some residents and large co-operations are impacting the sole power provider, JPS.
To tackle this problem, JPS partnered with World Bank, A data science firm (The Impact Lab), and Energy Sector Management Assistance. But that is the boring stuff and you can read more about that here. The interesting part is that they are using Machine Learning and Open Source tools to predict and identify perpetrators. This is a game-changer when compared to their previous efforts which costs over 10million. But let’s talk about the project.
Machine Learning!? Machines can’t learn…
Don’t worry, it is a concept in computing. Here is a video explaining what is machine learning. Essentially is it a new approach used to solve real-world problems. Instead of having Software Engineers write all the rules of the application, we instead use data and algorithms to predict what those rules are. We train the application to accomplish various tasks with the help of additional data.
This approach is similar to how humans learn. Throughout our lives we have different experiences (data) of which impacts (model) our perception. These experiences shape how we react (prediction) to future events. The more we practice the better we become at that task. Think of a baby learning to walk.
The same approach was used on this project. Instead of writing all the rules to identify instances of electricity theft, they use the data & algorithms to determine the rules. JPS and its partners would create models from their existing data to train the system to understand what is considered as a non-technical loss.
The JPS-Handoff Project
Well according to their GitHub repository, that is the name of the project. After skimming over their documents and a few code snippets, this is what I’ve learned thus far.
It is an open-source project running on Python’s machine learning toolset and Django framework. It seems to provide three important functions:
- Consumes and process data from multiple formats.
- Create machine learning models based on the data and generate predictions based on those data set.
- Provide a user interface allowing the machine learning results/predictions to be used. This allows JPS to:
- Obtain an audit of their results.
- A details view of potential accounts flagged as potential culprits.
- Categorize or group results.
Essentially helping JPS staff to avoid tedious task of manually combining through thousands of data to understand and track down offenders. Additionally, integrating this application with their Advanced Metering Infrastructure (AMI) will provide real-time analysis on usage.
What Are The Benefits?
JPS currently pass those losses onto their paying customers. This has been put on the front burner; thanks to COVID-19 and stay at home orders. After a number of residents complained about the drastic increase in their electricity bill, JPS explain their pricing strategy.
This open-source project has the potential to help you avoid those unfair monthly charges. By allowing JPS to quickly identify offenders may put a stop to this practice. It also has the potential of preventing future losses even if consumers change their behavior.
This is a perfect example of how open-source can help us solve challenging problems.
If you are interested in seeing an example of the application in action, comment below and I may dedicate a future post to it.
Like the post above? Check out these related posts: