کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

Adaptive Physics-Aware Deep Learning for PMU-Aided Power System State Estimation With Topology-Based Partitioning

نویسندگان :

SeyedHamed MirMohammadAli Roudaki¹ Mehrdad Abedi²

1- دانشگاه صنعتی امیرکبیر(پلی تکنیک تهران) 2- دانشگاه صنعتی امیرکبیر(پلی تکنیک تهران)

کلمات کلیدی :

Power system state estimation،PMU،physics-aware neural networks،deep learning،LSTM

چکیده :

Power system state estimation (PSSE) is a fundamental function of energy management systems, providing essential situational awareness for monitoring, control, and security assessment of modern power grids. The increasing penetration of renewable energy sources and the growing complexity of power system dynamics pose significant challenges to conventional weighted least squares (WLS)–based state estimation methods in terms of robustness, scalability, and real-time performance. Although phasor measurement units (PMUs) enable high-accuracy and time-synchronized measurements, their widespread deployment is constrained by installation and communication costs, which necessitates optimal PMU placement strategies. In this paper, a physics-aware deep learning (PANN) framework is proposed for real-time power system state estimation by integrating optimal PMU placement with topology-aware neural network modeling. The proposed approach exploits the physical structure of the power network by partitioning the system into observable subgraphs, each associated with at least one optimally placed PMU. Long short-term memory (LSTM) networks are employed to capture temporal correlations among system states, enabling online learning and adaptive estimation under dynamic operating conditions. Simulation results on IEEE 14-bus and 30-bus systems demonstrate superior accuracy and robustness compared with conventional PMU-based and purely data-driven estimators.