Original title: Transformer-Based Solar Irradiance Forecasting Model for Coastal and Microclimate-Sensitive Areas of First District of Batangas
Jeffrey S. Sarmiento, Dylan Josh Domingo Lopez, Gerard Francesco DG. Apolinario · September 2026
Predicting how much sunlight hits solar panels is vital for running solar power systems well. In coastal areas like Batangas District 1 in the Philippines, ocean breezes, sudden clouds, and shifting humidity make sunlight patterns change rapidly. Traditional forecasting methods struggle to track these fast shifts. To solve this, researchers developed a specialized computer model named FAT-Former. The model uses a deep learning structure called a Transformer, which tracks patterns over time to predict solar irradianceThe amount of sunlight energy hitting a surface, like a solar panel., the power of sunlight hitting a surface. In tests, FAT-Former clearly beat a basic persistence model. It hit the lowest average error during peak daylight hours and retained 101.30% of the actual observed variance. Even so, the model struggled during rain and on days with rapidly shifting sunlight, and its predicted uncertainty ranges were under-calibrated. Overall, the study proves that tailored models can handle tricky coastal weather, but they still need improvements for sudden storms.
Planning a picnic, you check the forecast once at breakfast. But near the coast, clouds roll in and out so fast that your morning forecast is stale by noon. So you keep checking the sky and moving. In this paper, the forecasting model does the same—it watches recent sunlight and weather changes and updates its forecast, instead of sticking to one early prediction.
Solar power is hard to rely on when weather changes fast, especially near coasts. Better forecasts help grid operators balance supply and demand, avoid outages, and make solar energy cost-effective. This model brings that reliability to places like Batangas.
The model performs worse when it rains or when daylight changes wildly, and its uncertainty estimates are too narrow, meaning they don't cover enough possible outcomes. It was only tested in one district, so we don't know if it works in other coastal places.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Transformer-Based Solar Irradiance Forecasting Model for Coastal and Microclimate-Sensitive Areas of First District of Batangas — Jeffrey S. Sarmiento, Dylan Josh Domingo Lopez, Gerard Francesco DG. Apolinario — Energies — doi:10.3390/en19184448
Abstract licence: CC BY
Explanation generated by Synthesis from the abstract — it is not the original paper.