Jung-Hyun Yang, Sang-uk Park, Journal of Internet Computing and Services, Vol. 25, No. 6, pp. 9-21, Dec. 2024
10.7472/jksii.2024.25.6.9, Full Text:
Keywords: TimesNet, solar power generation prediction, short-term prediction, Time Series Data, Energy Management, dimension transformation

Abstract

This study proposes the TimesNet model for short-term prediction of solar power generation and compares it with LSTM, GRU, and Transformer models. The model was trained and evaluated using solar power generation data from Kangwon National University's Samcheok campus from 2020 to 2023. The TimesNet model effectively captures complex time series patterns through a unique approach that transforms one-dimensional time series data into two-dimensional image format. Experimental results show that the TimesNet model outperformed other models in various evaluation metrics including MSE, MAE, and NMAE. The findings of this study are expected to play a crucial role in ensuring power grid stability, increasing economic efficiency, and optimizing energy storage systems. Future research can extend this approach to various renewable energy sources, contributing to energy system optimization for achieving carbon neutrality.

Statistics
Cite this article