Abstract
COVID-19 has directly affected the refined petroleum products industry due to significant changes in energy usage patterns induced by local and global responses to the pandemic. With its onset and persistence, these products faced systematic shifts in their demand patterns, which need to be discovered and monitored by oil producers to respond appropriately to market needs. We perform a mid-term (annual scale) time series mining using a machine-learning-based approach of matrix profile to primarily unveil the post-COVID-19 behavior of the demand generation processes of four petroleum products, including gasoline, distillate fuel oils, kerosene-type jet fuel, and propane. The results indicate that some refined products have responded robustly to the pandemic, while major changes in the demand patterns for other products are evident. This study's outcome will help refineries lay out short to medium-term plans and make prediction-based adjustments to their production and refining processes.