Ultra short-term wind power prediction based on an improved temporal convolutional network

Ultra short-term wind power prediction is becoming increasingly crucial for optimizing power plant operations, conserving energy, and enhancing grid stability. Accurate forecasts of wind power output enable operators to strategically adjust energy production methods, thereby improving efficiency and reducing operational expenses. However, accurately predicting wind power requires addressing the complex relationships between various meteorological factors and wind power generation, as well as accounting for both spatial and temporal characteristics of wind power data. To address these challenges, this study proposes an enhanced prediction scheme incorporating an improved temporal convolutional network (TCN). Initially, partial least squares (PLS) is employed to capture the relationship between wind power generation and meteorological data, effectively reducing data dimensionality. To enhance the prediction model’s adaptability, the TCN framework incorporates a spatial convolutional layer and the convolutional block attention module (CBAM). This integration enables the model to capture both temporal and spatial features inherent in wind power data. Experimental results demonstrate that the proposed wind power prediction methodology outperforms existing approaches in terms of prediction performance. This approach has the potential to significantly enhance wind power forecasting accuracy, contributing to more efficient and reliable wind energy utilization.