Professor Choi Yosun's Team at Pukyong National University Wins Excellence Award at Solar Power Generation Forecast Academic Competition

Announcement of 'Development of Ensemble Machine Learning Model' for Accurate Forecasting

Jung Dahee (left) and Kim Dawon, doctoral candidates, who won the Excellence Award at the Solar Power Generation Forecast Academic Competition.

[Asia Economy Yeongnam Reporting Headquarters Reporter Dongguk Lee] Kim Dawon and Jung Dahee, doctoral students in the Department of Energy Resources Engineering at Pukyong National University, along with Professor Choi Yosoon, received the Excellence Award, the Korea Photovoltaic Society President's Award, at the Korea Institute of Energy Research (KIER) Solar Power Generation Forecast Academic Competition.

This competition, held alongside this year's Korea Solar Energy Autumn Academic Conference, was organized to quantitatively verify current solar power forecasting technologies and to improve overall technological capabilities.

The Pukyong National University team presented a paper titled "Development of an Ensemble Machine Learning Model for Solar Power Generation Forecasting," which received excellent evaluations.

In this paper, they presented research results on forecasting technologies aimed at improving the accuracy of renewable energy generation predictions, including solar and wind power.

Currently, a renewable energy generation forecasting system has been introduced in the electricity market, where solar and wind power operators predict and submit renewable energy generation amounts one day in advance and receive settlement payments if the actual generation is within a certain error margin (8%) on the day. However, solar power generation, which is affected by weather changes, faces difficulties in power supply and planning due to meteorological variability.

The Pukyong National University students conducted correlation and principal component analyses using numerical weather prediction (NWP) data and solar power generation data collected from 15 solar power plants from January to December 2020.

Based on this, they trained five machine learning models (kNN, SVR, RF, DNN, XGB), validated them using annual data from 2021, and selected and proposed an ensemble machine learning-based solar power generation forecasting model based on the evaluation metric nMAE for quantitative analysis.

This content was produced with the assistance of AI translation services.