On the basis that the wind power prediction error obeys normal distribution

an algorithm for calculating the minimum sample size with the given confidence level and maximum permissible error is proposed

as is a method of computing the confidence level under the condition of the given error range and sample size. Then an analysis is carried out according to the typical significance level and the error range. It is demonstrated that the statistical confidence level of the sample is low when the sample size is small

which cannot represent the overall data. Finally

an assessment algorithm for wind power prediction is proposed by taking into account the rationality of sample size of the day-ahead wind power prediction accuracy rate. The actual examples show that the proposed method is more scientific and reasonable

avoiding unreasonable assessment fines.