Why Conventional SCADA Cannot Catch Inter-Area Oscillations in Time

Inter-area oscillations — electromechanical oscillations in the 0.1–1 Hz frequency band involving geographically separated generator groups — represent one of the most consequential threats to large-scale grid stability. Left undetected or undamped, these oscillations can cascade into blackouts. Conventional SCADA systems, operating at much lower update rates, simply cannot provide the time-tagged, high-resolution data needed for modal analysis of such fast-moving phenomena.

PMU-based wide-area monitoring systems (WAMS) resolve this gap by providing GPS-synchronized phasor measurements — voltage magnitude, voltage angle, current magnitude, and frequency — from geographically distributed substations simultaneously. This time-tagged data stream makes modal analysis achievable at grid scale. The dataset examined here spans more than 50 sources, including active patents across US, EP, CN, IN, and AU jurisdictions and peer-reviewed academic literature, covering the period from 2008 to 2025.

The institutional landscape is broad. On the patent side, the most active assignees include State Grid Corporation of China, China Electric Power Research Institute, NEC Laboratories America, ABB Research, General Electric Technology, Psymetrix Limited, Schweitzer Engineering Laboratories, Sandia National Laboratories (NTESS), Siemens, and Hitachi. Academic contributors include University of Naples Federico II, Aalborg University, Politecnico di Milano, Isfahan University of Technology, Iowa State University, and UNC Charlotte. Standards bodies such as IEEE and international organisations including IEC provide the measurement and communication standards — notably IEEE C37.118 — that underpin interoperable PMU deployment.

Seven Algorithms, One Goal: Extracting Modal Parameters from PMU Signals

Signal processing is the analytical core of PMU-based early detection — it transforms raw phasor time series into actionable modal parameters: oscillation frequency and damping ratio. A comparative evaluation of seven techniques applied to ringdown signals from wide-area measurement systems has been documented by Veer Surendra Sai University of Technology (2017), providing practitioners a structured basis for algorithm selection: Prony Analysis, FFT, S-Transform, Wigner-Ville Distribution, ESPRIT, Hilbert-Huang Transform, and Matrix Pencil Method.

Each algorithm targets the same output — oscillation frequency and damping ratio — but through different mathematical routes. The University of Naples Federico II presents an optimized Hilbert Transform (HT)-based method specifically designed to separate individual inter-area oscillatory components and extract characteristic parameters from PMU-acquired electrical signals. Isfahan University of Technology proposes an augmented Prony method incorporating intrinsic mode functions (IMFs) derived from synchrophasor data, providing online monitoring capability with an energy- and phase-based indicator to identify the most influential generators for specific modes.

Woosuk University’s DFT-based method with an exponential window function handles segmented transient PMU data to capture time-varying modal characteristics of interarea, regional, and subsynchronous modes during faults. For near-resonance conditions — where nonlinear transient dynamics distort PMU waveforms — GEIRI North America developed an Extended Prony Analysis method grounded in normal form theory, validated on Kundur’s two-area system and the IEEE 39-bus system.

“Multi-band oscillation classification systems enable categorical early detection by distinguishing inter-area, local, subsynchronous, and other oscillation types from the moment of detection — enabling fine-grained early classification rather than simple threshold alarms.”

On the patent side, Schweitzer Engineering Laboratories’ foundational US patent (2009, active) combines a real-time modal analysis module with decision and control logic that processes signals from intelligent electronic devices (IEDs) to detect undesirable oscillations and activate remedial actions in real time. General Electric Technology’s active EP patent (2022) classifies oscillations across five predefined frequency bands — each attributed to uniquely different source types — and determines magnitude, phase, and damping characteristics for each detected frequency. Siemens’ active Indian patent (2024) introduces dynamic adaptation of the number of oscillation variable samples used for parameter estimation, addressing the critical early-detection problem of providing accurate damping parameters rapidly after oscillation onset.

Pinpointing the Disturbance: Forced Oscillation Source Localization

Knowing that a poorly damped mode exists is necessary but insufficient — grid operators also need to know which generator or controller is injecting periodic energy into the network before they can take corrective action. Forced oscillation source localization is operationally distinct from modal identification, and it has attracted a dense cluster of patents and publications.

State Grid Corporation of China holds the most prolific patent portfolio in this area. Its US patent (2014, active) and related Indian filings (2015, 2023) share a common algorithmic core: computing the algebraic sum of energy flow directional factors across regions of the interconnected grid to determine the disturbance source position online in real time, while reducing the computational burden associated with integration-based energy function methods.

The Anhui Electrical Power Research Institute of State Grid introduces an equivalent electrical distance approach, mapping grid nodes onto a complex plane and solving a Geiger-theory-based seismological optimization problem combined with the half-plane method to locate the source using PMU-monitored frequency variance waveforms — validated on the IEEE 30-bus system. IIT Delhi contributes a dissipating energy-based technique that continuously updates network and load information using SCADA measurements, with the notable capability of handling multiple concurrent disturbance sources with different activation times — a realistic operational scenario.

A resonance-specific extension from a 2020 publication compares mode shapes of forced and natural oscillations in the actual US Eastern Interconnection to identify the source as the location exhibiting the largest angular difference between the two mode shapes. Al-Mostajed Technologies’ two-step synchrophasor signal processing approach characterizes forced oscillations by frequency, duration, nature, and source location using PMU data within a WAMS framework, with explicit handling of white noise in PMU data. General Electric Company’s pending US patent (2023) represents the frontier: a machine learning model first generates a candidate set of power system components, followed by an optimization model that identifies the specific source component, asset type, and controller type.

The real-world complexity of source localization is illustrated by Hohai University’s analysis of a Chinese grid (2016), which documents inter-area oscillations driven by random wind power fluctuations — a general forced oscillation (GFO) mechanism involving narrow-band random excitations rather than pure periodic disturbances. This is a materially different problem from single-source periodic injection, and it underscores the importance of algorithm diversity in operational WAMS deployments. According to IEC, interoperability standards for synchrophasor data exchange are foundational to enabling multi-vendor WAMS architectures that can apply these diverse localization algorithms across grid boundaries.

From Detection to Action: Wide-Area Damping Control Architectures

PMU-based wide-area damping controllers (WADCs) are the active counterpart to passive monitoring — they translate real-time modal intelligence into power system control commands that suppress oscillations before they reach dangerous amplitudes. Detection alone cannot prevent instability; the control loop must close fast enough to matter.

Hydro-Québec/IREQ presents a model-free system identification approach using PMU-derived modal participation factors to select appropriate synchronous generators for control through a WADC algorithm demonstrated to suppress multiple critical inter-area modes simultaneously across the IEEE 68-bus and 145-bus test systems, including under PMU noise and missing PMU scenarios. UNC Charlotte’s measurement-based MIMO transfer function identification approach similarly enables selection of the most observable input and most controllable output for WADC design, integrating a discrete linear quadratic regulator with Kalman filtering for robust damping.

Sandia National Laboratories holds an active US patent (2021) for a PMU-based control system explicitly targeting inter-area oscillation damping to protect against catastrophic blackouts, receiving phasor measurements from two or more AC transmission line locations to generate power control commands. NEC Laboratories America discloses a complementary patent family for grid-scale storage control, in which PMU-derived oscillation mode amplitudes, phases, frequencies, and damping coefficients are used in modal analysis to identify critical modes, and a power oscillation dampening control signal directs storage resources — coupling real-time detection with non-generator damping resources.

Generator redispatch offers an alternative damping mechanism when continuous modulation is impractical. Iowa State University demonstrates that synchrophasor-derived dynamic data can supply a modal sensitivity formula to rank the optimal generator pairs for redispatch to improve damping of interarea modes, validated on the New England 10-generator system. Keio University extends the control design paradigm by decomposing a large-scale power system into a network of passivity-short subsystems and using data-driven matrix inequalities to design wide-area controllers without requiring a full nonlinear system model.

ABB Research holds two complementary active European patents addressing system reliability in wide-area control contexts. The 2012 patent introduces a phasor-aligning unit that time-aligns phasor sets from geographically separated measurement groups — a critical synchronization layer underpinning any effective WADC. The 2020 patent addresses measurement failure resilience by replacing failed actual PMU measurements with model-estimated values, ensuring control continuity even under PMU outage conditions. Standards from IEEE on synchrophasor measurement (C37.118) and the broader framework provided by NERC reliability standards shape the operational requirements these control architectures must satisfy.

EPRI’s Texas Grid case study (2022) proposes a two-dimensional scanning method to identify grid zones critical to forced oscillations and then uses active power modulation through inverter-based resources (IBRs) as a damping actuator — a novel control pathway made feasible by the real-time situational awareness provided by WAMS.

Renewable Integration and the Machine Learning Frontier

Increasing renewable penetration is making continuous PMU-based monitoring non-negotiable. Research from the University of Tennessee (2017) shows that increasing PV penetration monotonically decreases damping of dominant inter-area oscillation modes in the US Eastern Interconnection, while also potentially spawning new oscillation modes under inappropriate inverter parameter settings. Wind power integration similarly drives inter-area oscillations through random wind power fluctuations — a general forced oscillation mechanism documented in a real Chinese grid by Hohai University (2016).

Machine learning is increasingly augmenting classical signal processing to meet this growing challenge. TERNA S.p.A. — Italy’s national grid operator — developed an artificial neural network trained offline on frequency measurements from actual PMU devices to enable real-time identification of inter-area electromechanical mode parameters. This approach reduces the latency of mode identification compared to iterative numerical methods while maintaining accuracy under real-world PMU noise conditions.

Sichuan University proposes a digital twin-based oscillation source localization system integrating generative adversarial imputation networks for missing data repair and super-resolution measurement techniques — addressing the data quality challenges that arise when PMU coverage is incomplete. General Electric Company’s pending US patent (2023) for a two-step ML plus optimization locator represents the commercial frontier: automated AI-assisted localization that identifies not just the source component but also the asset type and controller type responsible for the disturbance.

Psymetrix Limited’s active EP patent (2018) and parallel US patents offer a practically elegant approach requiring minimal additional instrumentation: using the phase relationship between grid frequency oscillations and active power oscillations on inter-tie lines to identify which grid subsystem contributes to cross-boundary oscillations. Hitachi’s active EP patent (2020) focuses on event localization for operator decision support, complementing oscillation detection with fault event attribution using high-resolution time-synchronized PMU data. As noted by IEA in its analyses of grid flexibility requirements, the increasing share of inverter-based generation is fundamentally altering the inertia and damping characteristics of large interconnected systems — making the PMU-based WAMS capabilities described here central to future grid reliability.