Power grid anaysis

Power grid anaysis

Video
Team name / Company name: Green-Team
Team leader: Ivan Anikin
Challenge: no. 9: Grid Control by drones – picture recognition
Problem: The problem we have decided to solve is the problem with the power line grid analysis. Right now people have to get to the top of the power line to be able to analyze the state of the power line, if it has some anomalies or if it has malfunctioned.
Solution: Our solution contains a plan of a drones system with a client server for the drone, from where the pictures of the grid system are being sent to the web service on Azure. The web service includes machine learning algorithm detecting and analysing the power grids. Another part of the soulition is an administrator dashboard, where the whole analysis process gets displayed and notificates of any threads. Our presentation: https://prezi.com/p/edit/nnbenwhl3ing/
Impact: You will be able to automate this whole proces only needing a drone and a few IT people. You will be able to save money for helicopter rides to the top of the power line and without the risk on health. The whole process will also be much more time efficient, requiring less emloyees, resulting in less wages for the company to pay.
Feasibility: The solution we found is very simple to implement and is not expensive regarding time or finances. The hosting of the web service will cost less than a 1.000$ per year, the drones with good quality cameras cost quite a pack of money, however their lifespend is more that a decade or two.
What you built: We created a Flask web application and connected it with the Azure machine Learning resource, trained custom models and used pretrained solutions. Unfortunately we haven't finished the visualistion part of the problem due to lack of time, eventhough we only needed to connect the parts together.
What you had before: Nothing, the knowhow only. We chose the topic on the place.
What comes next: We'd love to continue our journey with this project, cause it inspires and insterests all of us. We have passion for machine learning and solving interesting problems. We love to play with electronics and would like to improve the hardware part of the problem as well. Azure Web service we created runs and creates the analysis we want, however if given a week or two and resources, we'd be able to add the anomalies detection and whole frontend for admin and client.