Yihao Fang

Mr. Yihao Fang

orcid

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 November 2024 in Preprints.org
Verified
Preprint
Open access
PDF file
Yihao FangYihao Fang
Yihao Fang
JLJin Liu
Jin Liu
HJHaiyang Jiang
Haiyang Jiang
WCWenhao Chen
Wenhao Chen
During extreme space weather events, transient geomagnetic disturbances initiated by solar eruptive activities can induce geomagnetically induced currents (GICs), which have severe impacts on power grid systems and oil/gas pipelines. Observations indicate that GICs in power grids are characterized by large fluctuation amplitudes, broad frequency ranges, and significant randomness. Their behavior is influenced by several factors, including the sources of space weather disturbance, Earth's electrical conductivity distribution, the structural integrity and performance of power grid equipment, and so on. Consequently, predicting local grid GICs has been a highly challenging task. This paper presents a hybrid prediction approach using actual GIC data from power grids and deep learning techniques. We employ various technical methods, including Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Convolutional Neural Networks (CNN), Long Short-Term Memory Networks (LSTM), and Attention mechanisms, to investigate short-term prediction methods for local grid GICs. The study uses GIC monitoring samples from November 8th to November 10th, 2004, for model training and testing. The results are evaluated using the coefficient of determination R^2 , root mean square error (RMSE), and mean absolute error (MAE). Preliminary research suggests that the CEEMDAN-CNN-LSTM-Attention combined model significantly improves prediction accuracy and reduces the time delay in GIC prediction during geomagnetic storms compared to using LSTM neural networks alone.

8 March 2024
Verified
Conference Paper
Open access
JLJin Liu
Jin Liu
Yihao FangYihao Fang
Yihao Fang
The Geomagnetic Induced Currents (GICs) generated in power grid by large-scale solar events may lead to electrical equipment damage and widespread power outages. Consequently, it is significant to achieve effective GIC prediction and risk assessment for local grids. The Hilbert-Huang Transform (HHT) is used to analyze geomagnetic components and GICs monitoring data. By extracting the responses to storms, the instantaneous amplitudes are synthesized as equivalent GIC disturbance indices. A combined prediction model based on deep learning methods is proposed. Concretely, the observation data of geomagnetic components and GIC are decomposed by Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), and the Intrinsic Mode Functions (IMFs) are predicted by a combination model constructed by Convolutional Neural Network (CNN) and Long-Short-Term Memory (LSTM) with the attention mechanism. Then, the predicted sequences of IMFs are accumulated to obtain the final forecasting. In addition, the GIC potential risk of local power grid is evaluated comprehensively by considering the solar wind energy injection, geomagnetic disturbances, equivalent GIC indices and the forecasting based on observations.

1 January 2025 in Processes
Verified
Journal Article
Open access
PDF file
Yihao FangYihao Fang
Yihao Fang
JLJin Liu
Jin Liu
HJHaiyang Jiang
Haiyang Jiang
WCWenhao Chen
Wenhao Chen
During extreme space weather events, transient geomagnetic disturbances initiated by solar eruptive activities can induce geomagnetically induced currents (GICs), which have severe impacts on power grid systems and oil/gas pipelines. Observations indicate that GICs in power grids are characterized by large fluctuation amplitudes, broad frequency ranges, and significant randomness. Their behavior is influenced by several factors, including the sources of space weather disturbance, Earth’s electrical conductivity distribution, the structural integrity and performance of power grid equipment, and so on. This paper presents a hybrid prediction using actual GIC data from power grids and deep learning techniques. We employ various technical methods, including complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), convolutional neural networks (CNNs), long short-term memory (LSTM) networks, and attention mechanisms, to investigate short-term prediction methods for local grid GICs. The study uses GIC monitoring samples from 8 November 2004 for model training and testing. The results are evaluated using the coefficient of determination R2, root mean square error (RMSE), and mean absolute error (MAE). Preliminary research suggests that the combined CEEMDAN–CNN–LSTM–attention model significantly improves prediction accuracy and reduces the time delay in GIC prediction during geomagnetic storms compared to using LSTM neural networks alone.