An unsupervised method for predicting photovoltaic potential in Canada

Abstract

To mitigate the effects of global climate change caused by fossil fuel emissions, Canada needs to reach net-zero emissions as soon as possible. However, for a country that relies heavily on non-renewable resources to heat homes, fuel transportation, and support industries, renewable alternatives must be reliable, efficient, and effective. One of the front-runners in sustainable energy solutions is solar power. Our team analyzed the photovoltaic (PV) potential of geographical sites across the country using data from the Canadian Weather Energy and Engineering Datasets (CWEEDS). Using k-means clustering, an unsupervised machine learning model, we placed 564 locations into 5 clusters and then predicted the PV potential for each cluster using a range of irradiance and radiation variables. Through plotting our results on scatter graphs, we concluded that the PV potential in most of Canada is much higher than the world average (4.11-6.96 kWh/m2). Furthermore, the province of Alberta—known for its tar sands and oil production—has the highest PV potential in the country. The province has the potential to become the leader in solar energy production in Canada. These findings can aid governments in optimizing their shift towards solar power. By identifying solar power as a strong alternative to fossil fuels, administrations can start working towards setting up solar farms in places where they would optimally serve Canadians in order to take the first step in decreasing our national carbon footprint.