Research & Publications

AI Applications in Smart Grids

We aim to provide an AI innovation perspective in smart grid applications to improve situational awareness for power system monitoring and control. This awareness will ultimately allow for overall upgrades of operation performances in more renewable power systems while simultaneously enhancing energy reliability, resiliency, and power quality. Key applicational scenarios across modern power systems, including cyberattack detection, state estimation, voltage/reactive power optimization, and post-contingency emergency control, are explored to elaborate AI-based models and solutions.

Robust Renewable Integration and Monitoring in Active Distribution Networks. To capture the impacts of the heterogenous uncertainties from DER outputs, imprecise line parameters, and measurements with noises, I develop a framework of model-based and data-driven state estimation leveraging the heterogeneous measurements from supervisory control and data acquisition (SCADA) systems and the emerging PMUs. The proposed algorithms obtain the upper and lower bounds of state variables for better monitoring power grids under the coordinated impacts of multiple uncertainties. We capture the heterogenous uncertainties currently in distribution systems and realize 100x acceleration compared with the conventional methods in this field. Towards a 100% renewable power grid in the near future, where significant uncertainty produces a far-reaching influence on system operation, the proposed algorithms will be promising to uncertainty analysis and provide a foundation for distribution system control and decision-making.

Voltage and Reactive Power Control for Distribution Networks with Inverter-connected PV. A real-time adaptive volt-VAR optimization (VVO) paradigm is developed to support DSOs in effective voltage control via multi-agent deep reinforcement learning (DRL). This work is a novel attempt to apply such an innovative AI technique to VVO control on voltage-regulating devices in distribution systems. This method realizes power loss reduction and voltage regulation concurrently. Under time-varying operating conditions, this work enables a distribution network to self-learn with the “cognitive” function of system operation by mimicking the human mind.Furthermore, by assigning the global control variables to multiple agents with effective information exchange, the proposed method addresses the scalability issue and improves computational efficiency in larger-scale distribution systems, compared with single-agent DRL-based algorithms.

Semi-Supervised Learning Cyberattack Detection in Cyber-physical Distribution Systems. The high dependence of cyber-physical power systems on information technology increases vulnerability to malicious cyberattacks. Moreover, recent research on unobservable false data injection attacks (FDIAs) reveals the high risk of secure system operation since these attacks can bypass current bad data detection mechanisms. Considering the high dimensionality and correlated nature of power system measurements, I explore the advanced ML technique using PMU data in state estimation for cyberattack detection and mitigation. Specifically, autoencoders are integrated into an advanced generative adversarial network (GAN) framework, which detects abnormal measurements under unobservable cyberattacks by capturing the unconformity between anomalies and secure measurements. Because of the expensive labeling costs and potential missing labeled data in practical systems, this method only requires unlabeled data and a few labeled data from measuring instruments by leveraging a powerful data generation capability of GAN and thus is semi-supervised learning. 

Adaptive Emergency Control to Harness Grid Resilience in the Aftermath of Extreme Weather. The objective of this study is to develop a data-driven, deep learning-based solution to prevent the propagation of cascading failures when the grid is challenged by unexpected contingencies or combinational contingencies under uncertain environments.  The proposed end-to-end technology will be an online platform capable of evaluating and predicting grid conditions and selecting emergency control actions focused on load-shedding strategies and determining timing and boundaries for splitting the grid into self-sustained islands, as needed, to mitigate the propagation of cascading failures. This can be considered an enhancement of the last line of defense to prevent a widespread blackout. Both load-shedding strategies and the determination of islanding boundaries require intensive computation and are currently performed offline. With the rapidly increasing penetration level of renewables, additional uncertainties are introduced into the grid operation, making the computation even more demanding and difficult. By combining the merits of the advanced soft actor-critic DRL framework with the automatic entropy adjustment, our grid emergency control approach accomplishes sampling efficiency, scalability, and stability of the shedding policies compared to the state-of-the-art.

Graph-based Fault-line Location in Poorly Observable Distribution Feeders. Quick and accurate Fault location helps the utilities find and clear the fault events and accelerate system restoration; however, this is a challenging task as the bidirectional power flow due to the increasing penetration of DERs leads to the mal-trip or fail-to-trip of conventional protection devices and imprecise fault location. To mitigate the impacts, I propose a novel graph-based fault location algorithm in distribution networks by advanced state estimation techniques integrating PMU data. This faulted-line location method running in a decentralized manner has a lower computation cost and enables fast fault location within 15 milliseconds, which shows the application potential in larger-scale systems. Our approach considers the impact of DER penetration on distribution system operation, and its location performance is independent of fault types and fault impedances. Furthermore, the proposed algorithm is robust against high-level noises in measurements.




Journal Papers [Google Scholar]

[1]    Y. Zhang, X. Wang, J. Wang, and Y. Zhang, Deep reinforcement learning based volt-VAR optimization in smart distribution systems, IEEE Transactions on Smart Grid, vol.12, no.1, pp.361-371, Jan. 2021.

[2]    Y. Zhang, J. Wang, and B. Chen, Detecting false data injection attacks in smart grids: a semi-supervised deep learning approach, IEEE Transactions on Smart Grid, vol.12, no.1, pp. 623-634, Jan. 2021.

[3]    Y. Zhang, J. Wang, and M. Khodayar, Graph-based faulted line Identification using micro-PMU data in distribution systems, IEEE Transactions on Smart Grid, vol. 11, no. 5, pp. 3982-3992, Sept. 2020.

[4]    Y. Zhang and J. Wang, Towards highly efficient state estimation with nonlinear measurements in distribution systems, IEEE Transactions on Power Systems, vol. 35, no. 3, pp. 2471-2474, May 2020. 

[5]    Y. Zhang, J. Wang, and Z. Li, Interval state estimation with uncertainty of distributed generation and line parameters in unbalanced distribution systems, IEEE Transactions on Power Systems, vol. 35, no. 1, pp. 762 -772, Jan. 2020.

[6]    Y. Zhang, J. Wang, and J. Liu, Attack identification and correction for PMU GPS spoofing in unbalanced distribution systems, IEEE Transactions on Smart Grid, vol. 11, no. 1, pp. 762-773, Jan. 2020. 

[7]    Y. Zhang, J. Liang, Z. Yun, and X. Dong, A new fault-location algorithm for series-compensated double-circuit transmission lines based on the distributed parameter model, IEEE Transactions on Power Delivery. vol. 32, no. 6, pp. 2398-2407, Dec. 2017. 

[8]    Y. Zhang, J. Wang, and Z. Li, Uncertainty modeling of distributed energy resources: techniques and challenges, Current Sustainable/ Renewable Energy Report, vol. 6, no. 2, pp. 42–51, Jun. 2019.

[9]    Y. Chen, Y. Y, and Y. Zhang, A Robust SOCP-based state estimation in integrated electricity-heat system, IEEE Transactions on Smart Grid, vol.12, no.1, pp.810-820, Jan. 2021.

[10]  M. Cui, M. Khodayar, C. Chen, X. Wang, and Y. Zhang, Deep learning based time-varying parameter identification for system-wide load modeling, IEEE Transactions on Smart Grid, vol. 10, no. 6, pp. 6102-6114, Nov. 2019.

[11]   M. Khodayar, Y. Zhang, J. Wang, and Z. Wang, Deep generative graph distribution learning for synthetic power grids, IEEE Power Engineering Letters. (Under the 2nd review)

[12]  Y. Zhang, J. Liang, Z. Yun, and F. Zhang, Distributed parameter model based single-line fault location algorithm for series-compensated double-circuit transmission lines, Automation of Electric Power Systems, 38(9), 61-68, 2017. 

Conference Papers

[1] Y. Zhang, J. Wang, and Z. Li, Interval state estimation with measurement and network parameter uncertainty in unbalanced distribution systems. 2019 IEEE Power Engineering Society General Meeting, GA, Atlanta, pp. 1-5.

[2] Y. Zhang, J. Liang, and P. Wang, Mutual impedance parameter modeling and accurate location algorithm of angled space crossed transmission lines, 2016 China International Conference on Electricity Distribution (CICED), Xi'an, 2016, pp. 1-6.

[3] H. Li, Y. Gan, H. Liu, Z. Zheng, and Y. Zhang, Advanced fault location system for EHV transmission lines, 2015 5th International Conference on Electric Utility Deregulation and Restructuring and Power Technologies (DRPT), Changsha, 2015, pp. 999-1003.

[4] H. Liu, T. Jia, L. Mou, X. Zheng, Y. Zhang, and F. Zhang, Improved traveling wave based fault location scheme for transmission lines, 2015 5th International Conference on Electric Utility Deregulation and Restructuring and Power Technologies (DRPT), Changsha, 2015, pp. 993-998. 


Patents

[1] A fault-location algorithm for series-compensated double-circuit transmission lines based on distributed parameter models (No. CN105738769B)

[2] Three-dimensional generalized parameter calculation model for transmission lines considering spatial arrangement (No. CN105653760B)

[3] Faulted section identification and fault location for incomplete-journey double-circuit transmission lines (No. CN105929305B)

[4] Asynchronous fault location algorithm in series-compensated double-circuit transmission systems (No. CN106124927B)