After dealing with freezing in a lecture and sweating in another across the hallway in RCH, I decided to use ioT and data to help provide a more efficient method of monitoring the cooling/heating and average energy consumption of the air conditioners.

What it does

  1. The Arduino Uno with Shield and Temperature Sensors measure the temperature at regular intervals in different lecture rooms. (38 million temperature records / 24 hours)

  2. The data is collected in Google Cloud Platform's BigQuery where the data can be cleaned and massaged.

  3. The data is processed by Google Cloud Platform's Auto ML Engine (Machine Learning Engine) to find trends in the 38 million records. The trends will help figure out the relationship between vents in different classrooms to find the minimum energy required to maintain the temperature.

  4. Results are recorded and pushed to the React Application where the results will be able to be continually monitored and energy consumption can be tweaked as per the optimization.

How I built it

Hardware I built a temperature sensing robot using the Arduino Uno and the temperature sensors. I wrote a script that would take the analog signal from the Arduino and convert it to a 4 significant figure signal.

Software The data had to be cleaned using BigQuery and Python Scripts. The Auto ML Engine ran to find correlations between the energy, temperature, and time

Web The website was build on React JS. It is a simple 1 page application with a floor plan of RCH and the optimized data from the Auto ML Engine.

Challenges I ran into

Hardware is hard. Playing with the sensor until the correct temperature signal arrived was difficult.

A lot of the data processing and cleaning took a fair bit of time and required some manual assistance.

Accomplishments that I'm proud of

I have successfully completed a large multi-part application and model that will accurately collect, analyze, and optimize the energy efficiency of a building in a similar way that ecobee or nest does.

What I learned

ioT isn't as bad as I thought it was.

I absolutely love big data.

38 million records is a lot of records

GCP will forever be my go to cloud platform

What's next for Leef Sensor

I will use this for my coop interview at ecobee (if I get the interview) :)

Share this project: