Regional consumption modeling and forecasting with GRU: Understanding consumer behavior with deep learning

Today, the increasing energy demand has led to a necessary increase in production capacity, as well as a need for energy quality control and load forecasting due to consumers' irregular and fluctuating energy needs. Recently, the improvement of energy quality has been in the spotlight with the increasing deployment of smart grids and their expanding applications. Furthermore, developments in the field of artificial intelligence (AI) integrated into smart grids have contributed to enhancing energy quality. Despite the high quality of energy production, the consumption band, characterized by consumers' diverse and unstable situations, leads to imbalances in the grid. Various methods have been initiated to address these imbalances, one of which is to derive the consumption characteristics of the grid and regulate production accordingly. Additionally, adjusting production based on the extracted consumption pattern will enhance both energy production and consumption quality. Our study aims to provide a new solution to these types of problems. The study involves extracting and organizing the existing consumption patterns of a sample location, analyzing them, and then determining the current energy demand. Furthermore, the study aims to investigate the causes of irregularities and predict future energy needs in advance. Consumption forecasting considers various factors (such as weather conditions, location of use, date of use, renewable energy use, annual consumption, etc.). While there are many methods in the field of AI for prediction, we chose to use the Gated Recurrent Unit (GRU) method, which we believe is suitable for our dataset. The method we used for prediction has been proven to be suitable for our dataset based on the RMSE results we obtained.