Korean Journal of Computational Design and Engineering

Amid growing concerns about climate change and global energy security, renewable energy utilization is increasingly recognized as a crucial solution. This study aims to enhance the accuracy of solar power generation forecasting by using spatial data refinement techniques based on Delaunay Triangulation. This approach integrates climate data from geographically distributed observation stations to create enhanced datasets for predictive modeling. By using GridSearchCV and Gradient Boosting Regression, model hyperparameters were optimized to achieve reliable forecasting results. Two predictive models were compared: Model_1 combined data from three climate observation stations within each triangulated area, using a weighted approach for input generation, while Model_2 used data from the closest single observation station for prediction. The results demonstrated that the triangulated, multi-station approach (Model_1) provided more accurate forecasts than the single-station approach (Model_2), underscoring the effectiveness of Delaunay Triangulation for solar power prediction. These findings offer valuable insights for solar power plant site selection and operational planning, potentially boosting energy efficiency and economic returns in solar power projects.

키워드 Artificial intelligence, Delaunay triangulation, Gradient Boosting Regression, Photovoltaic power generation, Power generation prediction