[2K6-GS-2] Machine learning

[2K6-GS-2-01] A study of electricity demand forecasting with wide-area meteorological data using dimensionality reduction

〇Masaya Nakayama¹, Shoichi Urano¹ (1. Meiji University)

Keywords:Machine Learning, LSTM, Dimensionality Reduction, electricity demand forecasting, meteorological data

It is known that electricity demand is closely related to people's behavior and is particularly affected by weather data. For this reason, conventional studies of electricity demand forecasting by electric power companies using meteorological data often use only information on meteorological observation points corresponding to the demand points to be forecasted. Therefore, this paper aims to improve the accuracy of electricity demand forecasts for the following day by using meteorological data not only for demand points but also for the entire country, taking into account that in Japan the weather tends to change from west to east, driven by the prevailing westerly winds, while forecasting electricity demand for the region.