eCON | Devpost

eCON

SIVA SUBRAMANIAN PARI ARIVAZHAGAN SUDHEER RAJA BALASUBRAMANIAM THEETHARAPPAN GAJANAN WADEKAR

eCONS- An Electricity conservation application.

A machine learning driven application, trained with convolution neural networks to predict and proactively notify users when electricity consumption crosses a threshold.

Our application basically has a mobile app (iOS and Android) as the front end and supported by a convolution neural network trained model at the backend. The application initially asks the user the maximum budget for the monthly electricity bill, for example, $30. Based on this input a daily limit ($1.00/day)and an hourly limit ($0.04/hr) can be inferred and that will form the basis of the application. The application tracks the electricity consumption of the various devices connected to the grid. If the consumption is in such a manner that it will be crossing the allowed hourly limit of $0.04/hr soon , the application will proactively notify the user the list of all the devices that are heavily consuming electricity and provide an interface to SWIPE RIGHT to keep the devices on and SWIPE LEFT to turn off the devices.

This is because the daily limit is just a rough estimated rate at which if one consumes electricity one will be within their budget, there might be some days that require MORE/LESS power. We always put the USER in the control and devices are turned of only when the user decides so.

The list which the user is presented is every single time optimised to have only the relevant devices which can be turned off and we learn this through the convolution neural network model and every time a user make a decision of leaving a device on or turning it off we learn from that incident and the model evolves from that knowledge.

So in the future, say on a Friday the list of devices would not feature the microwave oven and the air conditioner because the user tends to have his friends over for dinner on every Friday and had specifically not turned off the oven and the air conditioner on Fridays.

Now if we implement this at scale, the buildings in college campus can be treated like devices operating inside a home which is the college. We can predict the electricity consumption of individual buildings inside a college campus and the college management can allocate funds accordingly. This helps the institution in a financial standpoint, as it predicts the investment that is required in the future, now, with pretty good accuracy and since it is a neural network as time progresses the accuracy only improves.

The challenges we wish to be judged for are : State Farm challenge - Conserving electricity is not only for the users own good, but it also contributes to the wellness of the planet as it reduces net CO2 emissions. Citi challenge - Our app aims to reduce the electricity bill and also predicts the future electricity usage and thus helping users on a financial standpoint of being to take an informed decision when they are investing their hard earned money.

JP Morgan Chase Challenge - our app is concerning home automation which is very helpful for people which physical disabilities as they can control the various devices and monitor their household electricity usage and thanks to the proactive suggestions can help reduce their electricity bill.

Best use of Open Source - We trained our model based on an open Source dataset from data.com. (https://openei.org/datasets/dataset/commercial-and-residential-hourly-load-profiles-for-all-tmy3-locations-in-the-united-states)

OIT Challenge - When implemented at scale this can be a great cost saver for the OIT and the university as well.