Uncertainty modeling in wind power generation prediction by neural networks and bootstrapping

Accurate short-term wind power forecasting with quantification of the associated uncertainty is crucial for the management of energy systems including wind power generation. On top of the inherent uncertainty in wind speed, it is necessary to account also for the uncertainty in the relationship between wind speed and the corresponding power production, typically described by a power curve whose characteristic parameters are not precisely known in practice. In this paper, we propose a novel approach to wind power forecasting with uncertainty quantification. The approach can be schematized in two steps: first, short-term estimation of wind speed Prediction Intervals (PIs) is performed within a multi-objective optimization framework worked out by Non-dominated Sorting Genetic Algorithm-II (NSGA-II); then, the uncertainty in wind speed and the uncertainty in the power curve are combined via a bootstrap sampling technique, thus obtaining wind power PIs with same coverage as the wind speed PIs.