⚡ electrical engineering

Response-Based Frequency Stability Assessment under Multi-Scale Disturbances in High-Renewable Power Systems

This paper proposes a unified response-based frequency stability assessment method for high-renewable power systems that classifies multi-scale disturbances, constructs a unified disturbance-power model for online identification and quantification, and derives analytical frequency-response models to evaluate stability limits for both step and slope-type disturbances.

Original authors: Jinhui Chen, Huadong Sun, Ping Wu, Baocai Wang, Bing Zhao

Published 2026-08-14

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jinhui Chen, Huadong Sun, Ping Wu, Baocai Wang, Bing Zhao

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the electrical grid as a giant, high-speed trampoline. When you jump on it, the fabric stretches and bounces back. In the world of electricity, this "bouncing" is called frequency. Normally, the grid hums at a steady rhythm, like a perfect heartbeat. But if you suddenly add a heavy weight (like turning on a massive factory) or remove a spring (like a power plant shutting down), the trampoline wobbles. If it wobbles too much or too fast, the whole system can crash, leading to blackouts.

For a long time, scientists knew how to handle big, sudden jumps on the trampoline. They had a simple rulebook: if the jump is a fixed size, they could predict exactly how the trampoline would react. But recently, the trampoline has changed. We've started using more wind and solar power, which are like bouncy, unpredictable springs that don't always hold their shape. Now, the "jumps" aren't just big and sudden; they are also weird. Sometimes the weight appears for a split second and vanishes. Other times, it creeps up slowly, like a heavy backpack being filled with sand grain by grain. The old rulebooks can't handle these new, tricky shapes. If we try to use the old "big jump" rules for a "slow creep," we might think the trampoline is safe when it's actually about to collapse, or vice versa. This is the puzzle this paper tries to solve: how do we keep the trampoline safe when the jumps are strange, fast, or slow?

The authors of this paper, a team of researchers from China, have built a new "smart sensor" system to watch the trampoline in real-time. Instead of guessing what kind of jump is happening, their method listens to how the generators (the springs of the trampoline) actually react. By measuring the electrical "shakes" of these generators, they can figure out exactly what kind of disturbance is occurring, even if no one told them beforehand.

They discovered that disturbances come in four distinct flavors, and each needs a different safety rule:

The Short-Term Glitch: Imagine someone jumping on the trampoline for a split second and then jumping off. This happens fast, like when a solar farm briefly loses power during a storm. The old rules would treat this as a permanent heavy weight, making the system panic unnecessarily. The new method sees the weight disappear and knows the trampoline will bounce back fine.
The Big Step: This is the classic scenario—a heavy weight is dropped and stays there. The old rules work okay here, but the new method adds a "safety margin" calculator to tell us exactly how close we are to the edge.
The Fast Ramp: Imagine the weight on the trampoline isn't just dropped; it's being added quickly, like a bucket of water pouring in over a few seconds. This is dangerous because the trampoline doesn't have time to adjust. The new method spots this rapid increase and predicts exactly how many seconds until the trampoline breaks.
The Slow Creep: This is the sneakiest one. The weight is added so slowly (over minutes) that the trampoline barely moves at first. The generators try to push back, but eventually, they get tired and run out of energy. Once they are exhausted, the trampoline starts to sink rapidly. The old rules often miss this because they assume the system can handle the slow creep forever. The new method realizes the generators are "running out of gas" and calculates exactly how much time is left before the crash.

The researchers tested their new "smart sensor" on a complex computer model of a real power grid, one that is heavily loaded with wind and solar power. They simulated all four types of weird jumps. The results showed that their method was spot-on. For the "Short-Term Glitch," it correctly predicted that the system would survive, whereas the old method would have falsely warned of a disaster. For the "Slow Creep," it correctly predicted the moment the generators would give up and the frequency would crash, while the old method thought the system was safe for much longer.

Crucially, the paper argues against using the old "step disturbance" rules for these new, weird events. If you try to explain a slow, creeping weight as a single big jump, you get the wrong answer. The authors show through their simulations that this old approach can be dangerously optimistic about slow problems or overly pessimistic about fast, short ones.

The team also checked if their method would break if the sensors were a little bit inaccurate—like if the "shakes" were measured with a tiny bit of noise. They found that even with these small errors, the method still correctly identified the type of disturbance and gave a safe, conservative warning. It didn't get confused; it just became slightly more cautious, which is exactly what you want when dealing with a trampoline that might collapse.

In short, this paper doesn't just say "we have a new tool." It provides a specific, mathematical way to listen to the grid, identify the shape of the trouble, and calculate the exact time we have to fix it before the lights go out. It turns a confusing mess of "weird jumps" into a clear, four-step guide for keeping our power grid stable in a world of renewable energy.

Problem Statement
In high-renewable power systems, the decline of synchronous inertia and conventional primary frequency regulation has rendered frequency stability more vulnerable to unanticipated active-power disturbances. These disturbances are increasingly diverse in time scale and temporal evolution, ranging from short-term deficits (e.g., during Low-Voltage Ride-Through or HVDC commutation failures) to time-varying slope disturbances (e.g., cascading trips or persistent supply-demand gaps). Existing frequency-stability assessment methods typically assume known step disturbances or rely on pre-defined disturbance forms. Consequently, they lack a unified, response-based framework capable of identifying unanticipated disturbance types in real-time and computing class-dependent stability margins, particularly for time-varying slope disturbances where conventional System Frequency Response (SFR) models may fail to capture cumulative growth or post-regulation exhaustion dynamics.

Methodology
The paper proposes a response-based quantitative frequency stability assessment method that infers disturbance power directly from measured generator electrical responses (electromagnetic power increments). The methodology proceeds through three main stages:

Disturbance Classification: The method distinguishes between short-term and permanent disturbances using an early-stage power threshold ( $\Delta P_{sh}$ ) and a frequency threshold ( $\Delta f_{sh}$ ). Permanent disturbances are further categorized into step, second-level slope, and minute-level slope disturbances based on the inferred disturbance power magnitude and its rate of change.
- Short-term: Rapid rise and recovery within a short interval.
- Step: Sudden increase followed by a constant level.
- Second-level slope: Linear increase with a large slope, reaching limits within seconds.
- Minute-level slope: Slow linear increase over minutes, exhausting primary regulation before causing rapid frequency decline.
Disturbance Intensity Quantification:
- For short-term and step disturbances, intensity is quantified by the maximum inferred disturbance power magnitude.
- For slope disturbances, intensity is characterized by the disturbance-power slope ( $k_s$ or $k_m$ ), estimated via least-squares fitting or derived from power-balance conditions at the exhaustion threshold.
Analytical Stability Criteria:
- Short-term: An inertial energy-accumulation criterion is derived to calculate the critical disturbance power based on kinetic energy and load frequency modulation, preserving the finite-duration recovery feature.
- Step: Critical disturbance power is determined by the minimum of constraints derived from steady-state frequency deviation (governor limits) and transient frequency deviation (SFR model with power-electronic participation). A safety-margin index is defined.
- Second-level slope: An improved SFR model in the Laplace domain is used to predict the time-to-limit violation ( $t_m$ ) for linearly growing disturbances.
- Minute-level slope: A post-exhaustion rotor-motion model is developed. Once primary regulation is exhausted, the frequency evolution is modeled as a first-order differential equation driven by the remaining imbalance, allowing for the prediction of the time to frequency-limit violation after the exhaustion instant ( $t_1$ ).

Key Contributions

Unified Classification Scheme: A response-based framework that identifies four distinct disturbance types (short-term, step, second-level slope, minute-level slope) using only measured responses, without prior knowledge of the disturbance form.
Class-Dependent Analytical Criteria: Derivation of specific stability margins for each class:
- Energy-based criteria for short-term events to avoid the excessive conservatism of step-equivalent models.
- Dual-constraint (steady-state and transient) critical power limits for step disturbances.
- Time-to-limit prediction models for slope disturbances, including a novel post-exhaustion model for minute-level slopes that accounts for the depletion of primary regulation reserves.
Quantitative Margins: The method provides quantitative outputs including critical disturbance limits, safety margins, and time-to-limit predictions, enabling operators to assess how long a system can withstand evolving disturbances.

Results
The proposed method was validated on the CSEE-FS frequency stability benchmark system (a realistic AC-DC hybrid system with 56.1% renewable penetration) under various operating conditions and disturbance scenarios:

Short-term (LVRT): The method correctly discriminated stability for 400ms LVRT durations while identifying violations for 500ms and 600ms durations. In contrast, a conventional step-equivalent SFR approach overestimated the frequency nadir and incorrectly predicted instability for the 400ms case, demonstrating the necessity of preserving finite-duration recovery features.
Step Disturbances: The analytical criteria accurately predicted stability margins and frequency deviations for step load increases, matching simulation results across different renewable penetration levels.
Slope Disturbances: For second-level and minute-level slopes, the proposed models provided accurate predictions of over-limit times (relative errors generally within 8%). The minute-level model successfully captured the continued frequency deterioration after primary regulation exhaustion, whereas the conventional SFR model (even with slope input) overestimated the remaining time margin and missed the impending limit violation.
Robustness: Sensitivity analysis using Monte Carlo simulations showed that measurement errors (voltage magnitude, phase angle, latency) and uncertainties in renewable regulation capacity did not alter the disturbance-type classification. While these uncertainties introduced minor deviations in stability margins, the predictions remained generally conservative, supporting their utility for early warning.

Significance
The paper claims that its primary significance lies in addressing the gap in real-time, response-based assessment for unanticipated, multi-scale disturbances in high-renewable systems. By moving beyond the assumption of known step disturbances, the proposed framework enables the identification of evolving disturbance types (specifically slope disturbances) that are increasingly prevalent in modern grids. The derivation of specific analytical criteria for each disturbance class allows for more accurate and less conservative stability assessments compared to traditional SFR-based methods. Furthermore, the ability to predict the "time-to-limit" for minute-level slope disturbances offers a critical tool for operators to implement corrective actions before primary regulation reserves are fully exhausted and frequency collapse occurs. The method's robustness against measurement errors and its conservative nature in the face of parameter uncertainty further enhance its practical applicability for online security assessment.

More like this