With rising energy demand, accurate photovoltaic power forecasting has become essential for improving utilization and grid stability. This paper proposes a forecasting system based on dual-pyranometer measurements to enhance prediction accuracy. A microcontroller collects temperature, humidity, irradiance from dual pyranometer, weather, wind speed, and wind direction, and transmits data via Wi-Fi to a database for storage and analysis. The dual-pyranometer configuration effectively captures shading effects by detecting non-uniform irradiance, improving sensitivity to local and short-term climatic variations. Experimental results show that the fuzzy neural network model, trained on two months of data, achieves high forecasting accuracy with an RMSE of 0.418 W. To further validate the model, data from April to May 2025 were used for training, and June 2025 data were applied for prediction. Results indicate that the forecasted values remain close to the actual values, with errors converging to an RMSE of approximately 0.465 W. Despite good overall performance, rapid weather fluctuations still introduce certain deviations, revealing limitations in handling abrupt changes. Its practicality and adaptability highlight strong potential for future solar power applications, offering robust technical support for energy autonomy and sustainable development.
Ren et al. (Mon,) studied this question.