Short-Term Wind Power Interval Prediction Based on TCN-DBG-PSMAT

With the increasing proportion of new energy access, the accurate prediction of wind power is increasingly important for the stable operation and flexible scheduling of the integrated energy system. In view of the traditional prediction model, it is difficult to fully explore the potential dependence between variables, which cause problems of low prediction accuracy of wind power under high dimensionality and large amount of data. In this paper, we propose a short-term wind power interval prediction model based on hybrid deep-learning. The model integrates a Temporal Convolutional Network (TCN) with a Deep Bidirectional Gated Recurrent Unit (DBG) and Prob-sparse Multi-head Self-attention mechanism (PSMAT). Based on this, Quantile Regression (QR) and Kernel Density Estimation (KDE) are combined to obtain interval predictions of future values at different confidence levels. The method is validated on the output dataset of a wind farm in Australia. The experimental results show that the network proposed in this paper has better performance and can more accurately characterize the uncertainty of wind power output compared to other prediction models, laying the foundation for subsequent research on integrated energy system scheduling.