Network Interdependency-Informed Power System Dynamic Trajectory Prediction Utilizing Black-Box Modeling of Inverter-Based Resources

Jul 16, 2026

Black-box modeling of inverter-based resources (IBRs) enables real-time power-grid operation despite proprietary control architectures. Existing ML online trajectory predictors either accumulate large errors when multiple surrogate models are used concurrently or overlook measurement uncertainty, reducing practical deployability. This work introduces a network interdependency-informed online ML approach: a modular spatiotemporal attention network for each IBR’s black-box model and a hybrid physics-informed loss using a decoupled linearized AC power-flow formulation to preserve physical consistency without costly iterative solving, improving accuracy and robustness.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

Introduction

Inverter-based resources and grid modernization

Black-box modeling motivation

Related work on ML trajectory prediction

Network Interdependency-Informed Power System Dynamic Trajectory Prediction Utilizing Black-Box Modeling of Inverter-Based Resources Sungjoo Chung, Student Graduate Member, IEEE, Ying Zhang, Senior Member, IEEE, Meng Yue, Hantao Cui, Senior Member, IEEE arXiv :2607 .05843v2 [ ee ss . SY] 8 Jul 2026 Abstract—Black-box modeling of inverter-based resources (IBRs) has become essential for real-time grid operation and control in the presence of proprietary electronic control architectures. Existing machine learning (ML)-based online dynamic trajectory prediction approaches using IBR black-box models either significantly accumulate prediction errors when multiple surrogates are simultaneously used or ignore measurement errors, limiting their deployment in practical grids. To address these limitations, this paper proposes a novel network interdependencyinformed ML algorithm for online dynamic trajectory prediction in IBR-integrated power systems. A modular spatiotemporal attention network (STAN)-based predictor for the black-box modeling of each IBR unit is first proposed. Utilizing past measurements, the proposed STAN can effectively capture and predict the spatiotemporal dynamics of IBRs by employing an attention mechanism to attend to the most pertinent features for trajectory prediction. Furthermore, a novel hybrid physicsinformed loss function that integrates a decoupled linearized AC power flow formulation is proposed. The proposed loss function effectively ensures physical consistency of predictions within network operation while avoiding the computational complexity of iterative power flow solving, thereby enabling efficient gradient backpropagation and overall improved prediction accuracy. Case studies on the IEEE 14-and WECC 179-bus systems demonstrate that the proposed method achieves significant accuracy enhancement and robustness against measurement errors, outperforming recent ML-based trajectory prediction methods. Index Terms—Dynamic trajectory prediction, inverter-based resources, black-box modeling, physics-informed neural networks, machine learning, data-driven. I. INTRODUCTION INVERTER-based resources (IBRs) are at the forefront of grid modernization to enhance system stability, enable selfsynchronization, and support reliable operation under lowinertia conditions [1] . As the penetration of variable renewable energy continues to grow, inverter-based generation is projected to contribute more than 50% of the total capacity at any given time in future grids [2] . Grid-forming invertershave garnered growing attention due to their ability to actively control voltage outputs, in contrast to grid-following inverters. This capability enables them to provide immediate ancillary services, such as reactive power support, to enhance grid resilience during contingencies [3] . Unlike synchronous generators (SGs) that provide abundant inertia to counter frequency deviations after disturbances, IBRs contribute less inertia, making grids increasingly susceptible to disturbances. S. Chung and Y. Zhang are with the Department of Electrical and Computer Engineering, Oklahoma State University, USA (email: [sungjoo.chung@okstate.edu](sungjoo.chung@okstate.edu); [y.zhang@okstate.edu](y.zhang@okstate.edu)). M. Yue is with Brookhaven National Laboratory, USA. H. Cui is with the Department of Electrical and Computer Engineering, North Carolina State University, USA. To this end, accurately predicting transient dynamics becomes more challenging. Another issue is the proprietary nature of IBRs, as original equipment manufacturers (OEMs) often restrict access to detailed control and design information of IBRs due to confidentiality requirements. This limits the feasibility of conventional numerical and model-based approaches for dynamic trajectory prediction, and thus, blackbox modeling approaches for IBRs have attracted significant interest from academia and industry [4] . This paper thus focuses on developing and utili

FAQs

What problem does the paper address in ML-based online trajectory prediction for inverter-based resources?

It targets two key limitations: error accumulation when multiple surrogate models are used together and the neglect of measurement errors, both of which hinder real-world deployment.

How does the proposed method model inverter-based resources in a black-box setting?

It uses a modular spatiotemporal attention network (STAN) predictor for the black-box modeling of each IBR unit, leveraging past measurements to capture spatiotemporal dynamics.

What role does the hybrid physics-informed loss function play?

It integrates a decoupled linearized AC power-flow formulation to enforce physical consistency of predictions while avoiding the computational burden of iterative power-flow solving.

Jul 16, 2026

Upload by Theodore