Jizhong Zhu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Jizhong Zhu has a total of 193 co-authors. Here are the top co-authors by number of shared publications:
- Yun Liu
South China University of Technology
16 shared publications - Hanjiang Dong
Department of Electrical and Electronic Engineering, The Hong Kong Polytechnic University, Hong Kong
13 shared publications - Zipeng Liang
The Hong Kong Polytechnic University
12 shared publications - Chi Yung Chung
Hong Kong Polytechnic University
11 shared publications Haohao Zhu
10 shared publications- Yixi Chen
South China University of Technology
10 shared publications
Latest Publications
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72  (4) , 10844-10857
VT-DeepSeek: A Visual-Guided and Parameter-Efficient Pretrained Transformer for IoT-Enabled Multimodal Photovoltaic Forecasting in Consumer Energy Systems
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72  (4) , 10844-10857
VT-DeepSeek: A Visual-Guided and Parameter-Efficient Pretrained Transformer for IoT-Enabled Multimodal Photovoltaic Forecasting in Consumer Energy Systems
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Smart Grid
- Vol. 17  (6) , 5999-6016
Robust Energy Management of Hybrid Microgrids via a Nonintrusive Load Monitoring Framework With Dual-Task Learning
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Smart Grid
- Vol. 17  (6) , 5999-6016
Robust Energy Management of Hybrid Microgrids via a Nonintrusive Load Monitoring Framework With Dual-Task Learning
- Top Cited Papers
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Transportation Electrification
- Vol. 12  (5) , 8202-8220
Uncertainty-Aware Approaches Toward Managing Hydrogen-Powered Vehicles and Renewable Energy Units in Microgrid–Transportation Networks
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Transportation Electrification
- Vol. 12  (5) , 8202-8220
Top Cited PapersUncertainty-Aware Approaches Toward Managing Hydrogen-Powered Vehicles and Renewable Energy Units in Microgrid–Transportation Networks
- Open Access
- 4 September 2026
- Published by Elsevier in SSRN Electronic Journal
Coupling flexibility forecasting and day-ahead scheduling for EV aggregators: a frequency-enhanced Transformer approach
Reliable forecasts of electric vehicle (EV) charging flexibility help aggregators schedule charging at lower cost while satisfying user energy requirements. Most existing forecasting methods estimate load trajectories or intermediate charging parameters, but these outputs do not directly provide the cumulative energy envelopes required by optimization models. This paper presents a scheduling-oriented EV flexibility forecasting framework based on a frequency-enhanced patch time series Transformer (FreqPatchTST). The framework predicts baseline load increments, upward and downward flexibility capacities, and aggregate power limits rather than cumulative energy boundaries, thereby reducing the deterministic growth trend in the target series. An adaptive frequency-enhancement module is embedded into PatchTST to capture nonstationary daily and weekly charging patterns, and an asymmetric loss is used to penalize forecast deviations that can increase real-time settlement costs. The predicted flexibility envelopes are then incorporated into a day-ahead coordinated charging model with real-time deviation settlement. Experiments on real charging station data from Hangzhou, China, show that FreqPatchTST achieves an MAE of 0.1536, an RMSE of 0.2142, and an R² of 0.9558. In the scheduling task, it obtains the lowest total operating cost among the tested models, \textyen48,683.1, and reduces the gap to the ideal-foresight benchmark to about 7%. These results suggest that forecasting physically interpretable flexibility envelopes can support lower-cost EV aggregator scheduling under uncertain charging behavior.
- 4 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessCoupling flexibility forecasting and day-ahead scheduling for EV aggregators: a frequency-enhanced Transformer approach
Reliable forecasts of electric vehicle (EV) charging flexibility help aggregators schedule charging at lower cost while satisfying user energy requirements. Most existing forecasting methods estimate load trajectories or intermediate charging parameters, but these outputs do not directly provide the cumulative energy envelopes required by optimization models. This paper presents a scheduling-oriented EV flexibility forecasting framework based on a frequency-enhanced patch time series Transformer (FreqPatchTST). The framework predicts baseline load increments, upward and downward flexibility capacities, and aggregate power limits rather than cumulative energy boundaries, thereby reducing the deterministic growth trend in the target series. An adaptive frequency-enhancement module is embedded into PatchTST to capture nonstationary daily and weekly charging patterns, and an asymmetric loss is used to penalize forecast deviations that can increase real-time settlement costs. The predicted flexibility envelopes are then incorporated into a day-ahead coordinated charging model with real-time deviation settlement. Experiments on real charging station data from Hangzhou, China, show that FreqPatchTST achieves an MAE of 0.1536, an RMSE of 0.2142, and an R² of 0.9558. In the scheduling task, it obtains the lowest total operating cost among the tested models, \textyen48,683.1, and reduces the gap to the ideal-foresight benchmark to about 7%. These results suggest that forecasting physically interpretable flexibility envelopes can support lower-cost EV aggregator scheduling under uncertain charging behavior.
- 1 September 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Smart Grid
- Vol. 17  (5) , 4697-4711
Quantal Response-Based Adversarial Reinforcement Learning for Smart Grid Attack-Defense Game With Bounded Rationality Evolution
Smart grids enhance power transmission efficiency through integrated cyber-physical systems, but simultaneously increase vulnerability to unexpected contingencies and malicious attacks. Adversarial reinforcement learning (ARL) has emerged as a promising technique for grid attack-defense games, however, existing methods assume attackers’ complete rationality, forcing defenders to engage in complex saddle-point optimization problems (i.e., solving for worst-case Nash equilibrium). Such an issue not only leads to over-conservative defenders, but also poses severe convergence challenges. To mitigate these problems, this paper proposes a novel quantal response-based ARL (QR-ARL) method. QR-ARL generalizes ARL into bounded rationality via entropy regularization, which helps reflect realistic attacker behaviors and reduce defensive conservatism. Building on this, a novel self-paced learning technique is then tailored to manage training complexity by adaptively evolving the attacker’s rationality. Additionally, to address the potential conflict between short-term control costs and long-term stability objectives, a parameter space perturbation technique is proposed to promote exploration and reduce the agent’s reliance on local reward signals. Numerical experiments on IEEE 39-bus and 118-bus systems confirm the effectiveness and advantages of the proposed method over existing ARL methods.
- 1 September 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Smart Grid
- Vol. 17  (5) , 4697-4711
Quantal Response-Based Adversarial Reinforcement Learning for Smart Grid Attack-Defense Game With Bounded Rationality Evolution
Smart grids enhance power transmission efficiency through integrated cyber-physical systems, but simultaneously increase vulnerability to unexpected contingencies and malicious attacks. Adversarial reinforcement learning (ARL) has emerged as a promising technique for grid attack-defense games, however, existing methods assume attackers’ complete rationality, forcing defenders to engage in complex saddle-point optimization problems (i.e., solving for worst-case Nash equilibrium). Such an issue not only leads to over-conservative defenders, but also poses severe convergence challenges. To mitigate these problems, this paper proposes a novel quantal response-based ARL (QR-ARL) method. QR-ARL generalizes ARL into bounded rationality via entropy regularization, which helps reflect realistic attacker behaviors and reduce defensive conservatism. Building on this, a novel self-paced learning technique is then tailored to manage training complexity by adaptively evolving the attacker’s rationality. Additionally, to address the potential conflict between short-term control costs and long-term stability objectives, a parameter space perturbation technique is proposed to promote exploration and reduce the agent’s reliance on local reward signals. Numerical experiments on IEEE 39-bus and 118-bus systems confirm the effectiveness and advantages of the proposed method over existing ARL methods.
- Open Access
- 20 August 2026
- Published by Elsevier in SSRN Electronic Journal
Coupling flexibility forecasting and day-ahead scheduling for EV aggregators: a frequency-enhanced Transformer approach
Reliable forecasts of electric vehicle (EV) charging flexibility help aggregators schedule charging at lower cost while satisfying user energy requirements. Most existing forecasting methods estimate load trajectories or intermediate charging parameters, but these outputs do not directly provide the cumulative energy envelopes required by optimization models. This paper presents a scheduling-oriented EV flexibility forecasting framework based on a frequency-enhanced patch time series Transformer (FreqPatchTST). The framework predicts baseline-load increments, upward and downward flexibility capacities, and aggregate power limits rather than cumulative energy boundaries, thereby reducing the deterministic growth trend in the target series. An adaptive frequency-enhancement module is embedded into PatchTST to capture nonstationary daily and weekly charging patterns, and an asymmetric loss is used to penalize forecast deviations that can increase real-time settlement costs. The predicted flexibility envelopes are then incorporated into a day-ahead coordinated charging model with real-time deviation settlement. Experiments on real charging station data from Hangzhou, China, show that FreqPatchTST achieves an MAE of 0.1536, an RMSE of 0.2142, and an R2 of 0.9558. In the scheduling task, it obtains the lowest total operating cost among the tested models, \textyen48,683.1, and reduces the gap to the ideal-foresight benchmark to 7.29%. These results suggest that forecasting physically interpretable flexibility envelopes can support lower-cost EV aggregator scheduling under uncertain charging behavior.
- 20 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessCoupling flexibility forecasting and day-ahead scheduling for EV aggregators: a frequency-enhanced Transformer approach
Reliable forecasts of electric vehicle (EV) charging flexibility help aggregators schedule charging at lower cost while satisfying user energy requirements. Most existing forecasting methods estimate load trajectories or intermediate charging parameters, but these outputs do not directly provide the cumulative energy envelopes required by optimization models. This paper presents a scheduling-oriented EV flexibility forecasting framework based on a frequency-enhanced patch time series Transformer (FreqPatchTST). The framework predicts baseline-load increments, upward and downward flexibility capacities, and aggregate power limits rather than cumulative energy boundaries, thereby reducing the deterministic growth trend in the target series. An adaptive frequency-enhancement module is embedded into PatchTST to capture nonstationary daily and weekly charging patterns, and an asymmetric loss is used to penalize forecast deviations that can increase real-time settlement costs. The predicted flexibility envelopes are then incorporated into a day-ahead coordinated charging model with real-time deviation settlement. Experiments on real charging station data from Hangzhou, China, show that FreqPatchTST achieves an MAE of 0.1536, an RMSE of 0.2142, and an R2 of 0.9558. In the scheduling task, it obtains the lowest total operating cost among the tested models, \textyen48,683.1, and reduces the gap to the ideal-foresight benchmark to 7.29%. These results suggest that forecasting physically interpretable flexibility envelopes can support lower-cost EV aggregator scheduling under uncertain charging behavior.
- 15 July 2026
- Published by Elsevier in Applied Energy
- Vol. 424
Distributed optimal power and gas dispatch using variational inequalities and cooperative games
- 15 July 2026
- Published by Elsevier in Applied Energy
- Vol. 424
Distributed optimal power and gas dispatch using variational inequalities and cooperative games
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Sustainable Energy
- Vol. 17  (3) , 2911-2925
EV Grouping and Distributed Coordination to Promote EV Flexibility Aggregation for Charging Station Operation
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Sustainable Energy
- Vol. 17  (3) , 2911-2925
EV Grouping and Distributed Coordination to Promote EV Flexibility Aggregation for Charging Station Operation