About

Professor Le Xie is the Gordon McKay Professor of Electrical Engineering at Harvard's John A. Paulson School of Engineering and Applied Sciences. His primary teaching area is Electrical Engineering. His research focuses on applied mathematics, data science, and science and engineering for ClimateTech, with particular emphasis on energy resources and energy systems, environmental science and engineering, and the development of sustainable power grids in the era of AI. Professor Xie is actively involved in advancing AI and machine learning applications for power systems and has contributed to the integration of AI technologies in energy and climate-related engineering fields. He is recognized for his leadership in these areas and is engaged in fostering innovative solutions for sustainable energy and climate challenges.

Research topics

• Computer Science
• Engineering
• Electrical engineering
• Business
• Geography
• Computer Security
• Engineering physics
• Economics
• Environmental economics
• Economy
• Reliability engineering
• Physics
• Operations research
• Industrial organization

Selected publications

• LLM-Based Adaptive Distribution Voltage Regulation Under Frequent Topology Changes: An In-Context MPC Framework

  IEEE Transactions on Smart Grid · 2025-06-27 · 4 citations

  articleSenior author

  This paper proposes a large language model (LLM) based adaptive inverter control for distribution voltage regulation under frequent topology changes. We leverage the ability of the LLM to perform in-context learning and create a topology-adaptive surrogate model for power flow calculation. The surrogate model is then integrated with a long short-term memory-based load forecaster and a model predictive control (MPC) scheme to achieve the optimal inverter control that adapts to frequent topology changes. Unlike many existing works that assume fixed-topology grids or require the knowledge of all possible topologies when training a model, the proposed in-context MPC method tackles the distribution voltage control problem under various topologies and adapts to unknown topologies with limited data requirement for fine-tuning. The effectiveness of our method is demonstrated on a modified IEEE 123-bus test system.

  PublisherDOI

• A Physics-Informed Graph Neural Network Framework for N-2 Contingency Screening: A Real-World Texas Grid Study

  IEEE Open Access Journal of Power and Energy · 2025-01-01 · 1 citations

  articleOpen accessSenior author

  This paper proposes a physics-informed graph neural network (GNN) framework for scalable and efficient AC power flow-based N-2 contingency screening in large-scale power systems. Formulated as a graph classification problem, the approach is specifically designed to identify critical N-2 contingencies that are likely to result in infeasible post-contingency AC power flow solutions. The integration of physics-based domain knowledge into the neural network architecture enhances the model’s capability to capture the underlying physical behaviors governing power flow, thereby improving classification accuracy. Comprehensive numerical experiments on the real-world Texas transmission network demonstrate that the proposed method achieves a 37-fold improvement in computational efficiency over conventional simulation-based N-2 contingency analysis techniques, underscoring its potential for operational deployment in real-time or near real-time security assessment.

  PublisherDOI

• An Econometric Analysis of Large Flexible Cryptocurrency-mining Consumers in Electricity Markets

  Proceedings of the ... Annual Hawaii International Conference on System Sciences/Proceedings of the Annual Hawaii International Conference on System Sciences · 2025-01-01 · 5 citations

  articleOpen accessSenior author

  In recent years, power grids have seen a surge in large cryptocurrency mining firms, with individual consumption levels reaching 700MW. This study examines the behavior of these firms in Texas, focusing on how their consumption is influenced by cryptocurrency conversion rates, electricity prices, local weather, and other factors. We transform the skewed electricity consumption data of these firms, perform correlation analysis, and apply a seasonal autoregressive moving average model for analysis. Our findings reveal that, surprisingly, short-term mining electricity consumption is not directly correlated with cryptocurrency conversion rates. Instead, the primary influencers are the temperature and electricity prices. These firms also respond to avoid transmission and distribution network (T&D) charges - commonly referred to as four Coincident peak (4CP) charges - during the summer months. As the scale of these firms is likely to surge in future years, the developed electricity consumption model can be used to generate public, synthetic datasets to understand the overall impact on the power grid. The developed model could also lead to better pricing mechanisms to effectively use the flexibility of these resources towards improving power grid reliability.

  PublisherOA PDFDOI

• PowerAgent: A Road Map Toward Agentic Intelligence in Power Systems: Foundation Model, Model Context Protocol, and Workflow

  IEEE Power and Energy Magazine · 2025-08-20 · 6 citations

  articleSenior author

  The operational resilience of electric power grids is facing growing challenges caused by aging infrastructure, increasing system complexity, and a rising frequency of extreme weather events. Traditional control paradigms, built around deterministic models and human-in-the-loop decision making, will become insufficient to manage the escalating demands on power grids. In response, recent advances in artificial intelligence (AI)—particularly the emergence of general-purpose AI agents capable of tool use, reasoning, and task orchestration—offer a new direction for enhancing grid flexibility and resiliency. This article introduces the concept of the Power Agent: an AI-enabled, context-aware assistant that leverages foundation models, standardized tool interfaces, and structured workflows to support grid operation and planning decisions. We discuss the conceptual architecture, implementation pathways, and system-level benefits of deploying Power Agents in power grid operations, with an emphasis on augmenting operator capabilities, improving situational awareness, and reducing operational bottlenecks.

  PublisherDOI

• Actionable Measures of Demand Side Resources as a Part of Virtual Power Plants: Case studies in Texas

  IEEE Electrification Magazine · 2025-03-01 · 1 citations

  articleSenior author

  Many regions such as Texas face escalating grid stress due to rapid population growth, industrialization, and extreme weather events. Virtual Power Plants (VPPs), aggregating distributed energy resources (DERs) such as smart thermostats, behind-the-meter photovoltaics (BTM-PV), and flexible loads, offer a scalable solution for enhancing grid flexibility and resilience. This study quantifies the peak demand reduction potential of demand-side resources in the Texas region using Monte Carlo simulations. Results indicate that widespread smart thermostat adoption could potentially reduce peak demand by up to 3.98 GW, while 10% BTM-PV penetration could potentially offset over 2 GW during peak hours. Key challenges include steep net-load ramp rates and distribution system constraints due to reverse power flow. Additionally, demand response (DR) effectiveness is highly sensitive to consumer participation, shaped by enrollment models, incentives, and communication strategies. These findings emphasize both technical and socio-economic prerequisites for effective VPP deployment and offer a possible framework for other regions targeting enhanced grid reliability and decarbonization.

  PublisherDOI

• A Review of Safe Reinforcement Learning Methods for Modern Power Systems

  Proceedings of the IEEE · 2025-03-01 · 43 citations

  reviewSenior author

  Given the availability of more comprehensive measurement data in modern power systems, reinforcement learning (RL) has gained significant interest in operation and control. Conventional RL relies on trial-and-error interactions with the environment and reward feedback, which often leads to exploring unsafe operating regions and executing unsafe actions, especially when deployed in real-world power systems. To address these challenges, safe RL has been proposed to optimize operational objectives while ensuring safety constraints are met, keeping actions and states within safe regions throughout both training and deployment. Rather than relying solely on manually designed penalty terms for unsafe actions, as is common in conventional RL, safe RL methods reviewed here primarily leverage advanced and proactive mechanisms. These include techniques such as Lagrangian relaxation, safety layers, and theoretical guarantees like Lyapunov functions to rigorously enforce safety boundaries. This article provides a comprehensive review of safe RL methods and their applications across various power system operations and control domains, including security control, real-time operation, operational planning, and emerging areas. It summarizes existing safe RL techniques, evaluates their performance, analyzes suitable deployment scenarios, and examines algorithm benchmarks and application environments. This article also highlights real-world implementation cases and identifies critical challenges such as scalability in large-scale systems and robustness under uncertainty, providing potential solutions and outlining future directions to advance the reliable integration and deployment of safe RL in modern power systems.

  PublisherDOI

• PowerAgent: A Roadmap Towards Agentic Intelligence in Power Systems

  2025-06-05

  preprintOpen accessSenior author

  The operational resilience of electric power grids is facing growing challenges due to aging infrastructure, increasing system complexity, and a rising frequency of extreme weather events. Traditional control paradigms, built around deterministic models and human-in-the-loop decision-making, are becoming insufficient to manage the escalating demands on power grids. In response, recent advances in artificial intelligence-particularly the emergence of general-purpose AI agents capable of tool use, reasoning, and task orchestration-offer a new direction for enhancing grid flexibility and resiliency. This article introduces the concept of the Power Agent: an AI-enabled, context-aware assistant that leverages foundation models, standardized tool interfaces, and structured workflows to support grid operation and planning decisions. We discuss the conceptual architecture, implementation pathways, and system-level benefits of deploying Power Agents in power grid operations, with an emphasis on augmenting operator capabilities, improving situational awareness, and reducing operational bottlenecks.

  PublisherOA PDFDOI

• Efficacy and safety of Chinese medicine compound for the convalescent COVID-19 patients: Protocol of a multi-centered, randomized, double-blinded, placebo-controlled clinical trial (Preprint)

  2025-08-03

  preprint

  <sec> <title>UNSTRUCTURED</title> ABSTRACT Background: Convalescent coronavirus disease 2019 (COVID-19) refers to a series of clinical syndromes in patients with COVID-19 infection that follow the relevant discharge indications but do not fulfill the criteria for a clinical cure, and these patients are discharged from the hospital with residual multifunctional deficits, including coughing, fatigue, and insomnia. Due to the prolonged convalescent COVID-19 infection, patients continue to experience symptoms or develop new symptoms after three months of infection, and some symptoms persist for over two months without any apparent triggers, which has a significant impact on the health status and quality of life of the population. Patients with convalescent COVID-19 lack a definitive pharmacological treatment. Traditional Chinese medicine (TCM) exhibits a distinct, synergistic effect on the treatment of convalescent COVID-19. However, there exists a limited number of clinical trials on TCM with lower evidence levels in convalescent COVID-19; therefore, randomized trials are urgently required. Methods: A multicenter, randomized, double-blind, placebo-controlled, phase II clinical trial was performed to evaluate the efficacy and safety of Shenlingkangfu (SLKF) granules in treating patients with convalescent COVID-19 and lung-spleen qi deficiency syndrome. Eligible participants were aged 18–75 years, had a confirmed or physician-suspected severe acute respiratory syndrome coronavirus 2 infection at least six months prior, and satisfied clinical criteria. Individuals with a history of severe pulmonary dysfunction or major liver and kidney illness or those on medications were excluded. Multicenter subjects satisfying all criteria were assigned (1:1) randomly into an intervention group and a control group. After a 2-day adjustment period, A total of 154 participants were randomly divided into an intervention group and a control group. The intervention group was given the SLKF granules orally once a bag, 16.9 g, twice daily, whereas the control group received the SLKF granule simulation at the same dosage. The trial was conducted over 14 days, with assessments performed at baseline and 14 days. Results: The primary outcomes were the therapeutic efficacy rate and total clinical symptom score. The secondary outcomes included the fatigue self-assessment scale, pain visual analog scale, Pittsburgh sleep quality index, mini-mental state examination, hospital anxiety and depression scale, TCM syndrome score, C-reactive protein, erythrocyte sedimentation rate, and interleukin-6. Three routine examinations, liver and kidney function tests, and electrocardiography were used as safety indicators. Conclusions:This study aimed to verify whether SLKF granules can significantly improve clinical symptoms, including fatigue, loss of appetite, cough, phlegm, and insomnia, in patients with convalescent COVID-19. For a comprehensive investigation, additional clinical trials with larger sample sizes and longer intervention periods are required.Clinical Trial Registration Center NCT1900024524, Registered on 26 January, 2024. </sec>

  PublisherDOI

• An economic analysis method for ship charging and swapping station in smart grid

  Frontiers in Energy Research · 2025-03-12

  articleOpen accessSenior author

  The reliable power supply and economic analysis of ship charging and swapping station are crucial for promoting the electrification of the shipping industry and achieving the dual carbon goals. This paper focuses on the development of an economic analysis method for ship charging and swapping stations within smart grid application scenarios. Firstly, the cost model is established by considering the construction, operation, maintenance, and equipment replacement of ship charging and swapping stations. Secondly, an operational model is defined, outlining the constraints for charging and discharging processes as well as backup power capabilities. Thirdly, an economic analysis framework is developed to minimize total investment and operational costs, incorporating factors such as thermal power unit operation, wind power curtailment, and deep peak shaving of thermal units. Finally, the proposed models are validated through a case study using modified IEEE 9-bus and IEEE 30-bus systems, and the results demonstrate significant improvements in economic efficiency and system performance when incorporating ship charging and swapping station.

  PublisherOA PDFDOI

• LILAD: Learning In-context Lyapunov-stable Adaptive Dynamics Models

  ArXiv.org · 2025-11-26

  preprintOpen accessSenior author

  System identification in control theory aims to approximate dynamical systems from trajectory data. While neural networks have demonstrated strong predictive accuracy, they often fail to preserve critical physical properties such as stability and typically assume stationary dynamics, limiting their applicability under distribution shifts. Existing approaches generally address either stability or adaptability in isolation, lacking a unified framework that ensures both. We propose LILAD (Learning In-Context Lyapunov-stable Adaptive Dynamics), a novel framework for system identification that jointly guarantees adaptability and stability. LILAD simultaneously learns a dynamics model and a Lyapunov function through in-context learning (ICL), explicitly accounting for parametric uncertainty. Trained across a diverse set of tasks, LILAD produces a stability-aware, adaptive dynamics model alongside an adaptive Lyapunov certificate. At test time, both components adapt to a new system instance using a short trajectory prompt, which enables fast generalization. To rigorously ensure stability, LILAD also computes a state-dependent attenuator that enforces a sufficient decrease condition on the Lyapunov function for any state in the new system instance. This mechanism extends stability guarantees even under out-of-distribution and out-of-task scenarios. We evaluate LILAD on benchmark autonomous systems and demonstrate that it outperforms adaptive, robust, and non-adaptive baselines in predictive accuracy.

  PublisherOA PDFDOI

Recent grants

• Microgrid Interconnections Control via Voltage Angle Droop Methods

  NSF · $400k · 2016–2021

• Look-Ahead Coordination of Variable Resources for Providing Electric Energy and Regulation Services

  NSF · $194k · 2010–2013

• A Cross-Domain Data-driven Approach to Analyzing and Predicting the Impact of COVID-19 on the U.S. Electricity Sector

  NSF · $335k · 2020–2023

• EAGER: A Dynamical Systems Approach to Modeling and Controlling Responsive Demand in Electric Power Systems

  NSF · $297k · 2015–2018

• Collaborative Research: CyberSEES: Coupon Incentive-based Risk Aware Demand Response in Smart Grid

  NSF · $667k · 2013–2018

Frequent coauthors

• P. R. Kumar

  Texas A&M University

  58 shared

• Tong Huang

  49 shared

• Xiangtian Zheng

  Texas A&M University

  47 shared

• Chanan Singh

  43 shared

• Dongqi Wu

  Zhejiang Energy Research Institute

  37 shared

• Xinbo Geng

  Cornell University

  34 shared

• Dileep Kalathil

  Mitchell Institute

  32 shared

• Marija Ilić

  30 shared

Labs

• Le Xie LabPI

Education

• Ph.D., Electrical and Computer Engineering

  Carnegie Mellon University

  2009