Load Forecasting Method Based on Discrete Wavelet Denoising and Inverse Attention Mechanism-NBEATS Model

The base of load is small, with high volatility and uncertainty. With the integration of renewable energy such as photovoltaic and wind power, the instability and noise characteristics of load become increasingly prominent, posing challenges to high-precision prediction. Therefore, the article proposes a load prediction method based on discrete wavelet denoising and inverse attention mechanism NBEATS model. Firstly, a load denoising and smoothing model based on discrete wavelet transform was established. Stable load fluctuation information was obtained by decomposing stable frequency subsequences and establishing a sample set; Then, a load prediction model based on reverse attention mechanism and NBEATS was constructed. The key features of multi-dimensional meteorological information were dynamically mined through feature dimension attention weight self-learning, and the multi-scale temporal changes of load were modeled using the NBEATS model to improve prediction accuracy. Finally, based on the actual load data of a city in China, a case study analysis was conducted, and the results showed that the prediction accuracy of the proposed method can reach 96.75%, which is 2.5% higher than traditional methods.