Semantic Segmentation of Solar PV Panels and Wind Turbines in Satellite Images Using U-Net

Abstract

Global mission is to reduce the carbon footprint by using “Renewable Energy resources”. It is important to speed up the development of Renewable Energy Resources like Solar, Wind, Hydro electric et al. Implementation of Renewable Energy helps to tackle the climate change issue, as most of the energy resources are currently fossil fuel based. Information on installed capacity of Solar PV Panels and Wind Turbines along with forecasted load can enable grid operators to ensure optimal and reliable operation of system. Deep learning framework is used here to detect the Wind Turbines and Solar PV Panels in Satellite images. The current work aims to remove the manual effort which is currently involved in surveying the renewable energy resources. Building-level or neighborhood-level information on Solar PV panels and Wind Turbines enable analysis of Solar PV panels and Wind turbines deployment. Carbon footprint and Payback period can be calculated using the Deep Learning model outcome approximately for the installed locations and proposed locations. Dataset was acquired from Google Maps (Satellite view) for this work.