Abstract: As the proportion of wind power generation in the power system continues to rise, its strong volatility affects the frequency stability of the power system. Therefore, accurate evaluation of the fluctuation range of wind power plays an important role in the stable operation and scheduling of the power system. Currently, interval forecasting mostly uses recurrent neural networks, which limits the depth of the model. Additionally, traditional interval forecasting adopts upper and lower limit prediction methods, which have the disadvantages of difficult adjustment of loss function hyperparameters and limited initialization methods. This paper proposes an interval prediction scheme based on probability density function parameter estimation to address these issues. By utilizing probability density distribution functions, the scheme provides both deterministic and interval prediction results. Meanwhile, a temporal Transformer network is proposed, which can enhance local feature extraction capabilities while retaining the Transformer's global perspective. Compared to the baseline model in a public dataset, the results show that both the model and the prediction scheme in this paper can provide better prediction accuracy for both interval and deterministic predictions.