HarGharSolar : Recognition of Potential Rooftop PhotoVoltaic Arrays Using Geospatial Imagery for Diverse Climate Zones.

Abstract

As solar energy has been recognized as an inexhaustible source of energy, the solar photovoltaic installation business has taken the lead in today's market. Nowadays, people are investing more in green energy due to its harmless and everlasting supply of energy and also its boundless applications. With the adaptation of solar panels on the building rooftops, people often fail to think of the total energy that will be generated from the solar panel and if the generated power is sufficient enough to fulfill the power requirements of the whole building. Different climate zones receive different amounts of sunlight and thus, solar energy generation varies in all regions. Artificial Intelligence has evolved to bring significant development in this field as it helps in detecting rooftops that have a potential for solar photovoltaic systems and also helps to efficiently detect how much energy can be generated using the solar panels. Latest Deep Learning models like YOLO, EfficientNet, VGG ResNet etc are able to detect rooftops using geospatial images of zones and models like U-Net, SegNet etc are used to configure the solar photovoltaic system for the consumer. An additional model for the calculation of power generated considering different parameters like climate, topography will be built using advanced AI techniques. The best performing models will be finetuned and integrated with the front end to act as a one stop destination for the end user.