Short-term net load forecasting based on a multi-channel transformer-bidirectional long short-term memory fusion model incorporating photovoltaic generation

Accurate load forecasting at the power source end is important for the optimal scheduling of Integrated Energy Systems (IES). However, the nonlinearity and uncertainty inherent in renewable energy generation pose significant challenges to accurate net load prediction in IES. This paper proposes a net load forecasting model that integrates Transformer and Bidirectional Long Short-Term Memory (BiLSTM) networks. First, input data is preprocessed to ensure completeness and usability. Then, the Transformer module encodes temporal features with positional encoding, while a gated convolutional layer is used to extract spatial features. The extracted features are then input into the BiLSTM network to learn temporal dependencies in both directions and generate the final point prediction results. After producing point forecasts, an adaptive bandwidth kernel density estimation (ABKDE) method is used to quantify prediction uncertainty and construct prediction intervals at multiple confidence levels. Experimental results demonstrate that the proposed model outperforms traditional time series forecasting models in terms of both point forecasting accuracy and interval prediction performance, confirming its effectiveness and robustness in complex IES scenarios.

Text 126097.pdf - Published Version Restricted to Repository staff only Download (12MB) Official URL or Download Paper: https://linkinghub.elsevier.com/retrieve/pii/S0378...

Actions (login required)

View Item