Complex-Valued Time Series Based Solar Irradiance Forecast

• Published in: arXiv.org 2022-11
• Main Authors: Voyant, Cyril, Lauret, Philippe, Notton, Gilles, Jean-Laurent Duchaud, Garcia-Gutierrez, Luis, Ghjuvan Antone Faggianelli
• Format: Article
• Language: English
• Subjects: Autoregressive models; Gaussian process; Irradiance; Model accuracy; Regression models; Statistical analysis; Time series
• ISSN: 2331-8422
• DOI: 10.48550/arxiv.2203.06701

This paper describes a new way to predict real time series using complex-valued elements. An example is given in the case of the short-term probabilistic global solar irradiance forecasts with measurement as real part and an estimate of the volatility as imaginary part. A simple complex autoregressive model is tested with data collected in Corsica island (France). Results show that, even if this approach is simple to set up and requires very little resource and data, both deterministic and probabilistic forecasts generated by this model are in agreement with experimental data (root mean square error ranging from 0.196 to 0.325 considering all studied horizons). In addition, it exhibits sometimes a better accuracy than classical models like Gaussian process, bootstrap methodology or even more sophisticated model like quantile regression. The number of models that it is possible to build by generating complex-valued time series is substantial. Indeed, by using exogenous or ordinal variables and computed quantities coupled with complex (or multi-complex) numbers, many studies and many fields of physics could benefit from this methodology and from the many models that result from it.