In order to improve the performance of wind power probability interval prediction, a wind power interval probability prediction method is proposed which is the combination of variable bandwidth hybrid sliding Gaussian kernel density estimation (VHSKDE(Gaussian)) and normal sliding exponential iteration (NSEI) based on critical weight method and anti-entropy weight method. The combination method is called VHSKDE(Gaussian)-NSEI for short. Firstly, the error is obtained by point prediction based on variational mode decomposition and long short-term memory neural network(VMD-LSTM). Then, VHSKDE(Gaussian) and NSEI are used to estimate the probability distribution of the prediction errors, and forecast interval is obtained under corresponding confidence probability. Finally, four objective weight assignment methods are used to weight the VHSKDE(Gaussian) link and the VHSKDE(Gaussian)-NSEI combination link twice to generate the final wind power prediction interval. The research results show that the excellent performances of PICP and PIAW can be compatible by using the proposed VHSKDE(Gaussian)-NSEI prediction model under different levels of confidence. The VHSKDE(Gaussian)-NSEI prediction model has higher reliability and accuracy than NSEI and VHSKDE(Gaussian), and provides an important reference for wind power probability prediction.