Tao Su
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Tao Su has a total of 12 co-authors. Here are the top co-authors by number of shared publications:
- Changjun Xie
School of Automation, Wuhan University of Technology, Wuhan, China
2 shared publications - Ying Shi
School of Artificial Intelligence, Wuhan University of Technology, Wuhan 430070, China
2 shared publications Changxi Yue
1 shared publication- Changxi Yue
China Electric Power Research Institute, Wuhan 430070, China
1 shared publication - Fan Sun
Xinjiang Electric Power Research Institute of State Gird, Urumqi 830000, China
1 shared publication Feng Zhou
1 shared publication
Latest Publications
- 22 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Systems
- Vol. 39 (2) , 3548-3559
DTformer: An Efficient Digital Twin Model for Loss Measurement in UHVDC Transmission Systems
Ultra-high Voltage Direct Current system (UHVDC) can realize long distance transmission, and its loss measurement has great significance for loss reduction and fault diagnosis. Currently, loss measurements can be performed by model-driven and data-driven approaches. However, poor operating conditions can lead to system sensing errors, which can significantly reduce the accuracy and speed of measurement. To deal with this problem, we propose a digital twin architecture and constructs a data-driven learning model. The model, DTformer, is embedded in the digital twin architecture. And it has three remarkable processes: encoder-decoder based random feature extraction, loss measurement transfer learning and knowledge distillation. These processes extract parameter information and topology features from the data of UHVDC. The model has good accuracy, response time and resistance to the missing data. The comparison experiment shows that DTformer outperforms popular feature extraction models with an average decrease of 38.80% in RMSE and 42.82% in MAE, and the best R2 value is closest to 1. The analyses of lightweight and missingness show that the DTformer also has significant response time improvement and missing resistance. Finally, we also implement the visualization of digital twin to enable operators to perform monitoring, computation, analysis and management of UHVDC.
- 22 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Systems
- Vol. 39 (2) , 3548-3559
DTformer: An Efficient Digital Twin Model for Loss Measurement in UHVDC Transmission Systems
Ultra-high Voltage Direct Current system (UHVDC) can realize long distance transmission, and its loss measurement has great significance for loss reduction and fault diagnosis. Currently, loss measurements can be performed by model-driven and data-driven approaches. However, poor operating conditions can lead to system sensing errors, which can significantly reduce the accuracy and speed of measurement. To deal with this problem, we propose a digital twin architecture and constructs a data-driven learning model. The model, DTformer, is embedded in the digital twin architecture. And it has three remarkable processes: encoder-decoder based random feature extraction, loss measurement transfer learning and knowledge distillation. These processes extract parameter information and topology features from the data of UHVDC. The model has good accuracy, response time and resistance to the missing data. The comparison experiment shows that DTformer outperforms popular feature extraction models with an average decrease of 38.80% in RMSE and 42.82% in MAE, and the best R2 value is closest to 1. The analyses of lightweight and missingness show that the DTformer also has significant response time improvement and missing resistance. Finally, we also implement the visualization of digital twin to enable operators to perform monitoring, computation, analysis and management of UHVDC.
- Open Access
A Frequency Decomposition-Based Hybrid Forecasting Algorithm for Short-Term Reactive Power
This paper proposes a new frequency decomposition-based hybrid reactive power forecasting algorithm, EEMD-LSTM-RFR (ELR), which adopts a strategy of frequency decomposition prediction after ensemble empirical mode decomposition and then data reconstruction to improve the prediction ability of reactive power. This decomposition process can compress the high frequency of reactive power and benefits the following separate forecasting. Long short-term memory is proposed for the high-frequency feature of reactive power to deal with the forecasting difficulty caused by strong signal disturbance and randomness. In contrast, random forest regression is applied to the low-frequency part in order to speed up the forecasting. Four classical algorithms and four hybrid algorithms based on different signal decompositions are compared with the proposed algorithm, and the results show that the proposed algorithm outperforms those algorithms. The predicting index RMSE decreases to 0.687, while the fitting degree R2 gradually approaches 1 with a step-by-step superposition of high-frequency signals, indicating that the proposed decomposition-predicting reconstruction strategy is effective.
Open AccessA Frequency Decomposition-Based Hybrid Forecasting Algorithm for Short-Term Reactive Power
This paper proposes a new frequency decomposition-based hybrid reactive power forecasting algorithm, EEMD-LSTM-RFR (ELR), which adopts a strategy of frequency decomposition prediction after ensemble empirical mode decomposition and then data reconstruction to improve the prediction ability of reactive power. This decomposition process can compress the high frequency of reactive power and benefits the following separate forecasting. Long short-term memory is proposed for the high-frequency feature of reactive power to deal with the forecasting difficulty caused by strong signal disturbance and randomness. In contrast, random forest regression is applied to the low-frequency part in order to speed up the forecasting. Four classical algorithms and four hybrid algorithms based on different signal decompositions are compared with the proposed algorithm, and the results show that the proposed algorithm outperforms those algorithms. The predicting index RMSE decreases to 0.687, while the fitting degree R2 gradually approaches 1 with a step-by-step superposition of high-frequency signals, indicating that the proposed decomposition-predicting reconstruction strategy is effective.