arXiv subjects

Zhirui Liang

Publications and source records attributed to Zhirui Liang.

13 recordsLinked to original sources

arXiv2026

Inference as Flexibility: Ramp Management for Transmission-Connected AI Data Centres

The rapid growth of large AI data centres introduces new operational challenges for power systems, including rapid ramping, oscillatory load behavior, voltage fluctuations, and supply-demand balancing impacts. For example, the Alberta Electric System Operator (AESO) has identified transmission-connected data centres (TCDCs) as large non-conforming loads that may need to limit their point-of-connection ramp rates. Existing mitigation approaches mainly rely on exogenous electrical resources, such as battery energy storage systems (BESS). This paper presents a proof-of-concept demonstration of a complementary software-defined mitigation layer: using flexible large language model (LLM) inference serving as endogenous TCDC flexibility to partially offset AI training power ramps. We consider a 150 MW TCDC with training, inference, and base-load components. A measured LLaMA-2-70B fine-tuning power profile is scaled to represent an aggregate training block, while measured LLaMA-3.1-70B inference power traces are used to model batch-size-dependent inference flexibility. Three strategies are compared: BESS-only mitigation, batch-size-only control, and coordinated batch-size plus BESS control. Simulation results show that the hybrid strategy reduces BESS discharge energy by 71% and peak discharge power by 51%, while maintaining near-complete compliance with a 10 MW/min ramp limit.

eess.SY↗

arXiv2026

OpenG2G: A Simulation Platform for AI Datacenter-Grid Runtime Coordination

AI's growing compute demand and new datacenter buildouts present major capacity and reliability challenges for the electricity grid, leading to multi-year interconnection delays for new datacenters and bottlenecking AI growth. To ease this strain, datacenters increasingly offer rapid power flexibility in response to grid signals, where the datacenter can increase or decrease its power consumption by adapting its workload in real time. In order to understand the impact of large datacenters on the grid and to facilitate the design of effective coordination strategies, we build OpenG2G, a simulation platform for AI datacenter-grid runtime coordination. We show that OpenG2G is capable of answering a wide range of coordination questions by allowing users to implement and compare various control paradigms (including classic, optimization, and learning-based controllers), and quantify how AI model and deployment choices affect datacenter flexibility and coordination outcomes. This versatility is enabled by OpenG2G's modular and extensible architecture: a datacenter backend driven by real measurements of production-grade AI services, a grid backend built on high-fidelity grid simulators, and a generic controller interface that closes the loop between them. We describe the design of OpenG2G and demonstrate its usefulness through realistic grid scenarios and AI workloads.

cs.LG↗

arXiv2026

GPU-to-Grid: Voltage Regulation via GPU Utilization Control

While the rapid expansion of data centers poses challenges for power grids, it also offers new opportunities as flexible loads. Existing power system research often abstracts data centers as aggregate resources, while computer system research focuses on GPU energy efficiency and largely ignores grid impacts. To bridge this gap, we develop a GPU-to-Grid framework that couples device-level GPU control with power system objectives. We study distribution-level voltage regulation enabled by LLM inference flexibility, using batch size as a data-center-side control knob that trades off GPU power consumption, inference latency, and token throughput. We first formulate the problem as an optimization problem and then realize it as an online feedback optimization controller, implemented by the data center operator using its own empirical GPU power-performance model and real-time measurements from both the GPU and grid systems. Our key insight is that reducing GPU power alleviates lower-voltage violations, while increasing GPU power mitigates upper-voltage violations; this challenges the common belief that minimizing GPU power is always beneficial to power grids.

eess.SY↗

arXiv2024

Prescribing Decision Conservativeness in Two-Stage Power Markets: A Distributionally Robust End-to-End Approach

This paper presents an end-to-end framework for calibrating wind power forecast models to minimize operational costs in two-stage power markets, where the first stage involves a distributionally robust optimal power flow (DR-OPF) model. Unlike traditional methods that adjust forecast parameters and uncertainty quantification (UQ) separately, this framework jointly optimizes both the forecast model parameters and the decision conservativeness, which determines the size of the ambiguity set in the DR-OPF model. The framework aligns UQ with actual uncertainty realizations by directly optimizing downstream operational costs, a process referred to as cost-oriented calibration. The calibration is achieved using a gradient descent approach. To enable efficient differentiation, the DR-OPF problem is reformulated into a convex form, and the Envelope Theorem is leveraged to simplify gradient derivation in the two-stage setting. Additionally, the framework supports distributed implementation, enhancing data privacy and reducing computational overhead. By proactively calibrating forecast parameters and prescribing optimal decision conservativeness, the framework significantly enhances cost efficiency and reliability in power system operations. Numerical experiments on an IEEE 5-bus system demonstrate the effectiveness and efficiency of the proposed approach.

eess.SY↗

arXiv2024

Electricity Market-Clearing With Extreme Events

Extreme events jeopardize power network operations, causing beyond-design failures and massive supply interruptions. Existing market designs fail to internalize and systematically assess the risk of extreme and rare events. Efficiently maintaining the reliability of renewable-dominant power systems during extreme weather events requires co-optimizing system resources, while differentiating between large/rare and small/frequent deviations from forecast conditions. To address this gap in both research and practice, we propose managing the uncertainties associated with extreme weather events through an additional reserve service, termed extreme reserve. The procurement of extreme reserve is co-optimized with energy and regular reserve using a large deviation theory chance-constrained (LDT-CC) model, where LDT offers a mathematical framework to quantify the increased uncertainty during extreme events. To mitigate the high additional costs associated with reserve scheduling under the LDT-CC model, we also propose an LDT model based on weighted chance constraints (LDT-WCC). This model prepares the power system for extreme events at a lower cost, making it a less conservative alternative to the LDT-CC model. The proposed market design leads to a competitive equilibrium while ensuring cost recovery. Numerical experiments on an illustrative system and a modified 8-zone ISO New England system highlight the advantages of the proposed market design.

eess.SY↗

arXiv2024

Learning with Adaptive Conservativeness for Distributionally Robust Optimization: Incentive Design for Voltage Regulation

Information asymmetry between the Distribution System Operator (DSO) and Distributed Energy Resource Aggregators (DERAs) obstructs designing effective incentives for voltage regulation. To capture this effect, we employ a Stackelberg game-theoretic framework, where the DSO seeks to overcome the information asymmetry and refine its incentive strategies by learning from DERA behavior over multiple iterations. We introduce a model-based online learning algorithm for the DSO, aimed at inferring the relationship between incentives and DERA responses. Given the uncertain nature of these responses, we also propose a distributionally robust incentive design model to control the probability of voltage regulation failure and then reformulate it into a convex problem. This model allows the DSO to periodically revise distribution assumptions on uncertain parameters in the decision model of the DERA. Finally, we present a gradient-based method that permits the DSO to adaptively modify its conservativeness level, measured by the size of a Wasserstein metric-based ambiguity set, according to historical voltage regulation performance. The effectiveness of our proposed method is demonstrated through numerical experiments.

eess.SY↗

arXiv2024

Revealing Decision Conservativeness Through Inverse Distributionally Robust Optimization

This paper introduces Inverse Distributionally Robust Optimization (I-DRO) as a method to infer the conservativeness level of a decision-maker, represented by the size of a Wasserstein metric-based ambiguity set, from the optimal decisions made using Forward Distributionally Robust Optimization (F-DRO). By leveraging the Karush-Kuhn-Tucker (KKT) conditions of the convex F-DRO model, we formulate I-DRO as a bi-linear program, which can be solved using off-the-shelf optimization solvers. Additionally, this formulation exhibits several advantageous properties. We demonstrate that I-DRO not only guarantees the existence and uniqueness of an optimal solution but also establishes the necessary and sufficient conditions for this optimal solution to accurately match the actual conservativeness level in F-DRO. Furthermore, we identify three extreme scenarios that may impact I-DRO effectiveness. Our case study applies F-DRO for power system scheduling under uncertainty and employs I-DRO to recover the conservativeness level of system operators. Numerical experiments based on an IEEE 5-bus system and a realistic NYISO 11-zone system demonstrate I-DRO performance in both normal and extreme scenarios.

math.OC↗

arXiv2024

Multi-Objective Transmission Expansion: An Offshore Wind Power Integration Case Study

Despite ambitious offshore wind targets in the U.S. and globally, offshore grid planning guidance remains notably scarce, contrasting with well-established frameworks for onshore grids. This gap, alongside the increasing penetration of offshore wind and other clean-energy resources in onshore grids, highlights the urgent need for a coordinated planning framework. Our paper describes a multi-objective, multistage generation, storage and transmission expansion planning model to facilitate efficient and resilient large-scale adoption of offshore wind power. Recognizing regulatory emphasis and, in some cases, requirements to consider externalities, this model explicitly accounts for negative externalities: greenhouse gas emissions and local emission-induced air pollution. Utilizing an 8-zone ISO-NE test system and a 9-zone PJM test system, we explore grid expansion sensitivities such as impacts of optimizing Points of Interconnection (POIs) versus fixed POIs, negative externalities, and consideration of extreme operational scenarios resulting from offshore wind integration. Our results indicate that accounting for negative externalities necessitates greater upfront investment in clean generation and storage (balanced by lower expected operational costs). Optimizing POIs could significantly reshape offshore topology or POIs, and lower total cost. Finally, accounting for extreme operational scenarios typically results in greater operational costs and sometimes may alter onshore line investment.

eess.SY↗

arXiv2023

Data-Driven Inverse Optimization for Marginal Offer Price Recovery in Electricity Markets

This paper presents a data-driven inverse optimization (IO) approach to recover the marginal offer prices of generators in a wholesale energy market. By leveraging underlying market-clearing processes, we establish a closed-form relationship between the unknown parameters and the publicly available market-clearing results. Based on this relationship, we formulate the data-driven IO problem as a computationally feasible single-level optimization problem. The solution of the data-driven model is based on the gradient descent method, which provides an error bound on the optimal solution and a sub-linear convergence rate. We also rigorously prove the existence and uniqueness of the global optimum to the proposed data-driven IO problem and analyze its robustness in two possible noisy settings. The effectiveness of the proposed method is demonstrated through simulations in both an illustrative IEEE 14-bus system and a realistic NYISO 1814-bus system.

math.OC↗

arXiv2023

A Physics-Informed Machine Learning for Electricity Markets: A NYISO Case Study

This paper addresses the challenge of efficiently solving the optimal power flow problem in real-time electricity markets. The proposed solution, named Physics-Informed Market-Aware Active Set learning OPF (PIMA-AS-OPF), leverages physical constraints and market properties to ensure physical and economic feasibility of market-clearing outcomes. Specifically, PIMA-AS-OPF employs the active set learning technique and expands its capabilities to account for curtailment in load or renewable power generation, which is a common challenge in real-world power systems. The core of PIMA-AS-OPF is a fully-connected neural network that takes the net load and the system topology as input. The outputs of this neural network include active constraints such as saturated generators and transmission lines, as well as non-zero load shedding and wind curtailments. These outputs allow for reducing the original market-clearing optimization to a system of linear equations, which can be solved efficiently and yield both the dispatch decisions and the locational marginal prices (LMPs). The dispatch decisions and LMPs are then tested for their feasibility with respect to the requirements for efficient market-clearing results. The accuracy and scalability of the proposed method is tested on a realistic 1814-bus NYISO system with current and future renewable energy penetration levels.

cs.LG↗

arXiv2022

Weather-Driven Flexibility Reserve Procurement: A NYISO Offshore Wind Power Case Study

The growing penetration of variable renewable energy sources (VRES) requires additional flexibility reserve to ensure reliable power system operations. Current industry practice typically assumes a certain fraction of the VRES power production forecast as flexibility reserve, thus ignoring other relevant information, such as weather conditions. To address this, probability- and risk-based reserve sizing models have been proposed, which use probabilistic VRES power forecasts that mostly rely on historical forecast and actual VRES power data for model training. Hence, these approaches are not suitable for planned or newly installed wind farms, where no or insufficient historical data is available. This paper addresses this caveat. First, we propose a weather-driven probabilistic forecasting method for wind power installations using publicly available weather data. Second, we apply the resulting probabilistic forecasts to a novel risk-based flexibility reserve sizing model that is compatible with the current reserve procuring pipeline used by US ISOs. Finally, we compare the risk-based reserve requirements to industry practice, state-of-the-art reserve procurement methods, and a weather-ignorant benchmark with respect to system cost and security. Our results are obtained from real-world data on a 1819-bus NYISO system model with both on- and projected off-shore wind power installations, which highlight the usefulness of weather information wind power forecasting and demonstrate efficiency gains from risk-aware reserve procurement.

eess.SY↗

arXiv2022

Inertia Pricing in Stochastic Electricity Markets

Maintaining the stability of renewable-dominant power systems requires the procurement of virtual inertia services from non-synchronous resources (e.g., batteries, wind turbines) in addition to inertia traditionally provided by synchronous resources (e.g., thermal generators). However, the pricing of inertia provision has not been studied in a stochastic electricity market, where the uncertainty characteristics of renewable energy sources (RES) are considered. To fill in this research gap, this paper formulates a chance-constrained stochastic unit commitment model with inertia requirements and computes equilibrium energy, reserve and inertia prices using convex duality. Numerical experiments on an illustrative system and a modified IEEE 118-bus system show the performance of the proposed pricing mechanism. By allowing new virtual inertia providers to contribute to system inertia requirements, the total operating cost reduces. Moreover, the proposed stochastic electricity market internalizes RES uncertainty, which yields additional cost reductions by co-optimizing energy, reserve and inertia procurement.

eess.SY↗

arXiv2022

Operation-Adversarial Scenario Generation

This paper proposes a modified conditional generative adversarial network (cGAN) model to generate net load scenarios for power systems that are statistically credible, conditioned by given labels (e.g., seasons), and, at the same time, "stressful" to the system operations and dispatch decisions. The measure of stress used in this paper is based on the operating cost increases due to net load changes. The proposed operation-adversarial cGAN (OA-cGAN) internalizes a DC optimal power flow model and seeks to maximize the operating cost and achieve a worst-case data generation. The training and testing stages employed in the proposed OA-cGAN use historical day-ahead net load forecast errors and has been implemented for the realistic NYISO 11-zone system. Our numerical experiments demonstrate that the generated operation-adversarial forecast errors lead to more cost-effective and reliable dispatch decisions.

eess.SY↗