Uncertainty analysis of wind power prediction based on Granular Computing

Abstract

Wind energy is supplying an increasing proportion of demand in the electrical grid. An accompanied problem is that the operational reliability of the power system is affected by the inherent uncertainty and stochastic variation of wind generation which also leads to the wind power forecasts of low accuracy. Therefore, the point prediction of wind power produced by a traditional deterministic forecasting model having a low level of confidence could not reflect the uncertainty of wind generation which could not meet the requirements for the safe operation of a power system. This paper aims to use the method of the non-parametric estimation to model the probability density distribution of the errors of wind power forecasts and determine the regression function based on the estimated point or deterministic wind power forecasts. The intervals of wind power predictions reaching a certain level of confidence can be employed by system operators to estimate the operation costs and the potential risks.