Hybrid Model Based on Temporal Convolutional Network and Attention Mechanism for Short-term Wind Power Forecasting

Wind power forecasting plays a key role in the overall task of curbing greenhouse gas emissions. To achieve this goal, there is a growing argument for leveraging sophisticated deep learning techniques to improve the accuracy of predictive models and contribute to a greener and more sustainable future. This paper proposes a hybrid forecasting model based on deep learning, harmoniously integrating temporal convolutional network (TCN) and attention mechanism. By considering historical information from various time scales, this model achieves adaptive weighted summation. Besides, by forecasting the difference instead of directly forecasting the power generation, this model can achieve efficient and high-quality forecasts. This research fills gaps in the literature by introducing a novel TCN hybrid model and conducting systematic comparisons among deep learning models for wind power forecasting. Rigorous validation on real datasets confirms the effectiveness of the hybrid model, leading to significant innovations and insights that can drive advancements in predictive modeling for renewable energy resources.

I. Introduction

As global energy demand rises, the globe is shifting toward alternate renewable energy sources to minimize greenhouse gas emissions. The increasing penetration of renewable energy in the power system provides significant environmental and economic benefits, but the intermittent and variable traits of renewable energy sources poses difficulties to the dependability and safe operation of power systems [1]. Wind energy is presently one of the most frequently used renewable energy supplies, and its share in energy systems is growing. By far, the entire global wind capacity had reached 744 GW, accounting for 7% of global power consumption [2]. The unpredictable nature of the wind, on the other hand, provides a considerable level of uncertainty in wind power generation. Meteorological elements such as temperature, humidity, boost, wind direction, and wind speed impact the wind generated by turbines. Addressing these challenges will allow wind energy to be produced and used more extensively [3].