Researcher · Engineer · Collaborator

Dr Yihong Zhou (周羿宏)

University of Oxford, Oxford, UK.

Building the control layer for flexible, AI-native power systems.

Future power systems will not be operated simply by adding more forecasting tools and control dashboards. They will need a new algorithmic layer that turns millions of grid-edge devices and AI-supporting data centres into "grid-intelligent" assets that operators can trust. This ambition forms the foundation of my research.

Flagship Work

Representative works I want to build on.

Optimization method preview
INFORMS Journal on Computing · 2026 Faster joint chance constraints

Strengthened Wasserstein-robust approximation for large-scale decisions under uncertainty, with 100x+ speedups.

 GradMAP method diagram
arXiv · 2026 GradMAP for grid-edge flexibility

Multi-agent proximal learning for 1,000 agents without parameter sharing, embedding three-phase AC load flow; trained within 15 min on a single non-data-centre GPU.

 Power tracking curves for grid-intelligent AI data centres
IEEE Transactions on Industry Applications · 2026 Grid-Intelligent AI Data Centres for Primary Response

Shows how AI data centres can become fast grid resources, with validation on real GPUs.
Current Agenda

Three questions driving the next stage.

Can AI infrastructure support the grid it depends on?

Designing power-aware workload control and flexibility products for AI/HPC data centres.

Can learning agents be made safe enough for energy systems?

Combining optimization structure, probabilistic safety, and multi-agent learning.

Can grid computation become AI-native?

Building batched, accelerator-ready power-flow and dispatch tools for modern AI ecosystems.

CV & Background

PhD in grid flexibility, MSc in AI, now building at the interface of both.

English CV 中文 CV

构建面向灵活性资源与 AI 基础设施的电力系统控制层。

未来电力系统不能只是简单叠加更多预测工具和控制面板。它需要一层新的算法基础设施，把数以百万计的电网边缘设备和支撑 AI 的数据中心，转化为运行者能够信任的“grid-intelligent”资产。这一目标构成了我研究工作的基础。

代表性工作

我最想持续往下推进的几条研究线。

Optimization method preview
INFORMS Journal on Computing · 2026 更快的联合机会约束近似

面向大规模不确定性决策的 Wasserstein 鲁棒优化方法，实现 100 倍以上加速。

 GradMAP method diagram
arXiv · 2026 面向电网边缘灵活性的 GradMAP

1000 个智能体、无参数共享训练，并嵌入三相交流潮流计算；仅用一张非数据中心级 GPU，15 分钟内完成训练。

 AI data centre power tracking curves
IEEE Transactions on Industry Applications · 2026 Grid-Intelligent AI Data Centres for Primary Response

展示 AI 数据中心如何成为快速电网资源，并在真实 GPU 上完成验证。
当前议题

下一阶段围绕三个问题展开。

AI 基础设施能否支撑其所依赖的电网？

研究面向 AI/HPC 数据中心的电力感知工作负载控制和灵活性产品。

学习型智能体能否安全进入能源系统？

结合优化结构、概率安全保证和多智能体学习方法。

电网计算能否成为 AI 原生工具？

构建批量化、可加速的潮流与调度工具，融入现代 AI 生态。

简历与背景

博士研究电网灵活性，硕士训练人工智能，现在连接二者。

English CV 中文 CV

News

Jul 16, 2026 I joined the Internetification of the Electric Grid panel at the IRTF SUSTAIN RG workshop The Internet meets the Electricity Grid: Technical Standards and Societal Challenges, hosted by the University of Passau.
Jul 07, 2026 Our preprint Decision-Focused Scenario Generation and Selection for Efficient and Robust Grid Dispatch is now available on arXiv.
Jun 24, 2026 I spoke on AI Data Centres: Power Burden or Future Grid Flexibility Backbone? during London Climate Action Week, presenting our latest work on data-centre grid intelligence and compute flexibility at The ZERO Institute, University of Oxford.
Jun 23, 2026 Our preprint Supervised Reinforcement Learning for the Coordination of Distributed Energy Resources is now available on arXiv.
Jun 11, 2026 I presented our work on Supervised Reinforcement Learning for the Coordination of Distributed Energy Resources at PSCC 2026 in Limassol, Cyprus.
May 16, 2026 I will serve as a Special Session Chair for IEEE EI2 2026 Special Session 02 on collaborative planning and operation of source-network-load-storage in new-type power systems.
May 16, 2026 I will serve as a Special Session Chair for IEEE I&CPS Asia 2026 Special Session 18 on AI-enabled optimization for integrated energy and transportation systems in smart cities.
May 16, 2026 Our paper Grid-Intelligent AI Data Centres for Primary Response is now available in IEEE Transactions on Industry Applications.
May 13, 2026 Our preprint JAX-Based Batched AC Power Flow for GPU Acceleration and AI Ecosystem Integration is now available on arXiv.
May 12, 2026 Our paper Strengthened and Faster Linear Approximation to Joint Chance Constraints with Wasserstein Ambiguity has been published online in INFORMS Journal on Computing.
Apr 27, 2026 Our preprint GradMAP: Gradient-Based Multi-Agent Proximal Learning for Grid-Edge Flexibility is now available on arXiv.
Feb 03, 2026 We hosted the first FleXEdge project workshop at the University of Oxford, with collaborators from Imperial College London, the University of Edinburgh (EPCC), and LIST (Luxembourg Institute of Science and Technology).
Jan 30, 2026 Our paper, “Independent Aggregators Securing End-User Wasserstein Distributionally Robust Flexibility through Bilevel Incentives,” has been published in Applied Energy and is now available on ScienceDirect.
Jan 28, 2026 We have two papers accepted for (Power Systems Computation Conference) PSCC 2026! See you in Limassol, Cyprus!
Jan 25, 2026 I created my first personal website!
Oct 10, 2025 Our preprint FICA: Faster Inner Convex Approximation of Chance Constrained Grid Dispatch with Decision-Coupled Uncertainty is available on arXiv.
Sep 18, 2025 I officially received my PhD degree from the University of Edinburgh.
Jan 20, 2025 Our preprint Strengthened and Faster Linear Approximation to Joint Chance Constraints with Wasserstein Ambiguity is now available on arXiv.