A Forecast That Remembers the Cycle express
A solar panel on a roof produces little current in the morning, a lot at noon, and stays silent at night. Wind doesn't blow randomly either: it's often stronger in the morning and evening than during the day. Energy providers need to know in advance how much energy will be available in an hour or three, so they can switch on backup plants or store electricity in batteries in time.
The usual forecasting method is 'as it is now, so it will be': if the sun is shining brightly, they assume it will be just as bright in an hour. But this fails when day turns to evening or when the wind dies down. The new formula adapts to the daily rhythm: it compares not just the current value with the past, but the current value with what it was at the same time of day yesterday or a few days ago. It's like an experienced gardener who knows that cucumbers should be watered not 'when dry' but every evening after sunset.
Tests on real data from solar and wind plants showed that the new method is more accurate than the old one, especially on horizons of several hours. This means we can keep fewer gas turbines in reserve and trust the renewable energy forecast more.
Original abstract
as written by the authors arXiv:2602.18949
Time series in energy systems, such as solar irradiance, wind speed, or electrical load, are characterized by strong diurnal and seasonal periodicities. Accurate forecasting requires accounting for time varying statistical properties that stationary or classical persistence models cannot capture. A family of analytical forecasting operators for cyclostationary processes is introduced, extending persistence through a closed form coefficient $$\tilde{\lambda}(t,\tau)=\tfrac{1}{2}\bigl(1+\rho(t,\tau)\bigr)$$, where $$\rho(t,\tau)$$ denotes the local correlation between the current observation and its phase aligned time lag ($$\tau$$). This formulation preserves periodic variance and covariance, achieving a symmetry induced reduction of effective degrees of freedom. The resulting operator defines a training free analytical limit of persistence under periodic non stationarity. Validation on synthetic cyclostationary signals and empirical renewable energy datasets demonstrates consistent accuracy gains over classical persistence, particularly at multi hour horizons. By embedding temporal symmetry into the prediction process, the framework provides a physically interpretable, reproducible, and computationally minimal baseline for forecasting periodic processes across energy and complex systems.
Solar energy reaching Earth in one hour exceeds the entire world's annual energy consumption — the problem is just knowing exactly when and how much will arrive.