GridWatch banner

GridWatch

4 devlogs
10h 33m 52s

Read for the expert mode password or check the GitHub README. GridWatch is a machine learning-based prediction tool for power outages across the US. It is currently powered by weather data, but in the future, we hope to train the model on historic…

Read for the expert mode password or check the GitHub README. GridWatch is a machine learning-based prediction tool for power outages across the US. It is currently powered by weather data, but in the future, we hope to train the model on historical data as well as integrating OpenCV to detect risks through public video cameras and automatically generate a listing on Terac for a qualified person in the area to take care of the risk before it evolves into an outage. Users can also input their email address and receive emails when there is an outage risk in the area they selected. Additionally, users can submit risk reports and see current risk reports (everywhere as well as in a specific zip code) in a table, which admins can manage.
Expert mode password: 12367. The reason it is not available on the site is because it is part of a Terac integration. After accepting the job on Terac, they get the password to expert mode and get access to the place to submit data for ML model training.

This project uses AI

Used Claude for some debugging in the backend, along with fixing frontend formatting. All content and ideas are original and not from AI. This includes color scheme, site components, actions, layout, and formatting.

Demo Repository

Loading README...

eastmohan22

Shipped this project!

Hours: 10.56
Cookies: 🍪 169
Multiplier: 15.96 cookies/hr

My friends an I built a website that uses a Random Forest ML model to predict power outages before they happen. Additionally, I made a place for users to submit reports and a place for admins to manage those reports. Users can also see the reports in a table. The hardest part was getting the data from the risk submission form to submit. I was trying to use a json file and js program, but I ended up switching to using a Vercel KV database, which is good because it updates automatically and instantly, and it is a lot easier than the method I was initially trying. I’m proud of the risk reporting and report viewing features because they were a challenge but I was able to overcome them. The ML model also seems to be pretty accurate because when it was raining, the percent risk in my area went up.

eastmohan22

I fixed adding reports, so now it shows in the table. It is expandable in the table so users can see images by clicking the entry, keeping the table more compact when collapsed. Additionally, I added an admin panel where anyone with the password can manage the entries in the table. To store data, I used a Vercel Upstash for Redis KV database, which I used Claude to help me properly implement.

Attachment
Attachment
Attachment
0
eastmohan22

I had to redo a lot of work such as creating everything on the report.html page because by .git got corrupted somehow and I hadn’t pushed yet. I also (hopefully) fixed index.html issues so the link never goes directly to index.html and goes to main.html instead. also added a button for report risk/outage to each page and fixed the format for the buttons. index.html was our old home page, but I stopped updating it and switched to main.html. I think I’m going to remove the map from this page and put it on a separate page to see current reports near the user. If possible, I’ll make it so they can be marked as fixed.

Attachment
0
eastmohan22

I made report.html. This page allows users to report power issues or risks that they see. I will add a way for them to upload photos. also, I need to add the link for it in the header for all other pages. I will then add a way for people to see active reports within a certain radius of them.

Attachment
0
eastmohan22

I worked mainly on the frontend. There is old work in the deprecated UI work, and then I used AI to help figure out how to implement the formatting I had planned. I also added integration with the Geocodio API so users can search for any location in the US. In the search bar, it shows suggestions as you type. Before, there were only a few locations that we manually added, all of which were in Virginia. The backend was worked on mainly by a couple of other people, but I helped debug it. These hours being logged are only hours that I worked, not including their hours. I also added a thumbnail to the tabs.

Attachment
2

Comments

oggythebuilder01

THATS AMAZING!!

eastmohan22
eastmohan22 8 days ago

Thanks!