Probabilistic Prediction Intervals of Wind Speed Based on Explainable Neural Network

With the rapid growth of wind power penetration into modern grids, speed forecasting plays an increasingly significant role in planning and operation electric energy systems. However, existing methods are modeled as black boxes, which very complicated cannot be written down explicitly due to complex fluctuation characteristics series. To this end, study proposes a novel direct method based on explainable neural network (xNN) for deterministic probabilistic forecasting. It can theoretically extract nonlinear mapping features speed, thereby providing clear explanation relationship between input output model. Then, uncertainties statistically synthesized via kernel density estimation method. Finally, we use data from real farms Belgium verify feasibility effectiveness proposed The simulation results demonstrate that it is not only able accurately non-stationary feature series but also superior other benchmark algorithms prediction accuracy. Therefore, has high potential practical applications