Wind power has been developed rapidly as a clean energy in recent years. The forecast error of wind power,
however, makes it difficult to use wind power effectively. In some former statistical models, the forecast error was usually
assumed to be a Gaussian distribution, which had proven to be unreliable after a statistical analysis. In this paper, a more
suitable probability density function for wind power forecast error based on kernel density estimation was proposed. The
proposed model is a non-parametric statistical algorithm and can directly obtain the probability density function from the
error data, which do not need to make any assumptions. This paper also presented an optimal bandwidth algorithm for
kernel density estimation by using particle swarm optimization, and employed a Chi-squared test to validate the model.
Compared with Gaussian distribution and Beta distribution, the mean squared error and Chi-squared test show that the
proposed model is more effective and reliable.
Wind Power Prediction Based on LSTM Networks and Nonparametric Kernel Density Estimation
