Communication Dans Un Congrès Année : 2024

A New Approach to Solar Irradiance Forecasting: Using Complex-Valued Time Series

Résumé

We present a novel and efficient approach for short-term probabilistic forecasting of global solar irradiance, leveraging complex-valued time series modeling. In this framework, the real part of the time series corresponds to the actual irradiance measurements, while the imaginary part encodes estimates of the volatility, allowing the model to capture inherent variability and uncertainties in solar radiation. By employing a complex autoregressive model, we are able to effectively represent the rapid fluctuations in irradiance, which are often challenging to predict with traditional methods. This approach was applied to data collected on Corsica Island (France), where solar irradiance is highly variable due to dynamic weather conditions. The results indicate that our model not only provides accurate deterministic and probabilistic forecasts, but it also does so with minimal computational complexity and limited data requirements. Specifically, the root mean square error (RMSE) of the forecasts ranges from 0.196 to 0.325 across various forecast horizons. This level of performance is competitive with, and in some cases surpasses, more traditional models such as Gaussian processes, bootstrap methodologies, and even advanced techniques like quantile regression. Additionally, the complex-valued nature of the model allows it to naturally incorporate both signal amplitude and volatility, giving it an edge in scenarios with high variability. The simplicity and efficiency of this method suggest that it could be particularly useful in real-time forecasting applications where computational resources are constrained. Moreover, this technique has the potential to be extended beyond solar irradiance forecasting, offering broad applicability in other areas of physics and environmental science that rely on time series data with rapidly fluctuating signals. In summary, this approach not only opens up new possibilities for solar energy forecasting but also provides a generalizable framework for other domains where accurate short-term predictions are crucial.

Fichier non déposé

Dates et versions

hal-04926470 , version 1 (03-02-2025)

Identifiants

  • HAL Id : hal-04926470 , version 1

Citer

Cyril Voyant. A New Approach to Solar Irradiance Forecasting: Using Complex-Valued Time Series. 11th International Conference on Energy, Sustainability and Climate Crisis (ESCC 2024 at Corfu, Greece), 2024, Corfu Island, Greece. ⟨hal-04926470⟩
66 Consultations
0 Téléchargements

Partager

  • More