Short-term load estimation based on improved DBN-LSTM
Abstract Aiming at the rapid change and low forecasting accuracy of short-term power load forecasting, a forecasting model based on the improved deep belief network and long short-term memory network is proposed. By combining deep belief network with long short-term memory network, the model gives f...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2025-07-01
|
| Series: | Sustainable Energy Research |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s40807-025-00192-w |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Aiming at the rapid change and low forecasting accuracy of short-term power load forecasting, a forecasting model based on the improved deep belief network and long short-term memory network is proposed. By combining deep belief network with long short-term memory network, the model gives full play to the advantages of deep belief networks in deep feature extraction, and enhances the modeling ability of long short-term memory network for time series data. Deep belief network extracts potential features from power load data by multilayer nonlinear transformation, so as to improve the understanding ability of long short-term memory network. The pruning algorithm is used to optimize the redundant structure of the model, reduce the complexity and training time of the model, and maintain or improve the forecasting accuracy. Through comparative experiments, the model achieved a forecasting accuracy of 90.02% during the training process, which was 4.62% higher than the RF-CNN. When compared with traditional LSTM, the designed method was 7.55% higher. In practical verification, the error between the weekday power load forecasting and the true value was 5–7 kW. The error between the holiday power load forecast and the true value was 7–8 kW. The designed model has obvious advantages in feature extraction and sequence learning, which can effectively capture potential features in power load data, providing a reliable solution for achieving high-precision power load forecasting. |
|---|---|
| ISSN: | 2731-9237 |