An improved teaching-learning-based optimization for extreme learning machine in floating photovoltaic power forecasting
Floating photovoltaic systems provide better land use and higher energy output through water cooling effects, but accurate power forecasting remains challenging due to complex environmental factors and measurement errors. This study presents an improved teaching-learning-based optimization algorithm...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
Oxford University Press
2025
|
| Subjects: | |
| Online Access: | https://umpir.ump.edu.my/id/eprint/45478/1/An%20improved%20teaching-learning-based%20optimization%20for%20extreme%20learning%20machine%20in%20floating%20photovoltaic%20power%20forecasting.pdf https://doi.org/10.1093/ce/zkaf042 https://umpir.ump.edu.my/id/eprint/45478/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Floating photovoltaic systems provide better land use and higher energy output through water cooling effects, but accurate power forecasting remains challenging due to complex environmental factors and measurement errors. This study presents an improved teaching-learning-based optimization algorithm with extreme learning machine for floating photovoltaic power forecasting. The method uses an adaptive teaching factor that adjusts the balance between exploration and exploitation during optimization, replacing fixed teaching factors with continuous, iteration-based adjustment. The research evaluated the approach using comprehensive real data from a floating photovoltaic installation at Universiti Malaysia Pahang Al-Sultan Abdullah, Malaysia. The proposed method achieved superior forecasting accuracy compared to benchmark algorithms including standard teaching-learning-based optimization with extreme learning machine, manta rays foraging optimization with extreme learning machine, moth flame optimization with extreme learning machine, ant colony optimization with extreme learning machine and salp swarm algorithm with extreme learning machine. The improved teaching-learning-based optimization approach demonstrated a root mean squared error of 7.81 kW and coefficient of determination of 0.9386, outperforming all comparison methods with statistically significant improvements. The algorithm showed faster convergence, enhanced stability, and superior computational efficiency while maintaining accuracy suitable for real-time grid integration applications. Phase current measurements were identified as the most important predictors for floating photovoltaic power forecasting. The system achieved high prediction accuracy with most forecasts falling within acceptable error tolerance, making the proposed approach a reliable solution for floating photovoltaic power forecasting that supports grid integration and renewable energy deployment. The methodology addresses unique characteristics of aquatic solar installations while providing practical implementation viability for operational floating photovoltaic systems. |
|---|