Applying Machine Learning to Study the Relationship Between Electricity Consumption and Weather Variables Using Open Data

As the effects of climate change are becoming more and more evident, the correlation between climate and consumption of energy has become a growing concern in the world. Researchers have been exploring possibilities to include weather influence on electricity consumption forecasts. However, no published approaches have been applied to produce accurate predictions of electricity consumption for the Netherlands using dynamic hourly based weather data. One of the reasons could be that those published approaches were developed with data of other countries. This limits their applicability for the case of the Netherlands. Therefore, this paper introduces a novel approach, which uses various weather variables (wind speed, temperature, air pressure, and humidity) measured at different locations in the Netherlands. With dedicatedly designed machine learning architectures, this paper presents a model that incorporates the influences of the dynamic weather variables on hourly-based electricity consumption using only publicly accessible data. The approach was implemented using a common laptop. The accuracy of the proposed approach is also discussed. Future work for developing a practically implementable model to predict the electricity consumption is discussed in the end.