In the summer after my freshman year, I got a family connection to a project in the nearby town of Andover. Anil Navkal was leading Energize Andover, an effort to help the town conserve resources through technology.
My task: Analyze a few years’ worth of power data from Andover High School and look for patterns that point to energy saving opportunities.
I was advised to do some basic statistical analysis on the dataset using Python, but I wound up down the rabbit hole of machine learning. I used Scikit-learn to develop a Python module for processing the Metasys sensor data, training a model, and making predictions about future energy trends. The model makes predictions on factors such as time of day, time of year, special events (loaded from a standard
.ical file), and outside temperature.
It returns predicted values paired with approximate variances for those values, which can be used to compute an anomaly threshold.
This was my first time using Python and virtual environments (with Anaconda). It was my first time using Hugo to make a blog and first time publishing a site on GitHub Pages. This was also my first time working extensively with Git and Linux.