Implementing General-Order Frequency Dynamic Response Model and Frequency Excursion Duration Criterion in Unit Commitment Problem
Mohammad Rajabdorri, Bo Zhou, Lukas Sigrist, Enrique Lobato · 2025
Read Implementing General-Order Frequency Dynamic Response Model and Frequency Excursion Duration Criterion in Unit Commitment Problem by Mohammad Rajabdorri, Bo Zhou, Lukas Sigrist, Enrique Lobato (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-02136.txt.
Introduction
Mohammad Rajabdorri, Bo Zhou, Lukas Sigrist, and Enrique Lobato embed frequency dynamics into the unit commitment problem with a general order model. Unit commitment schedules which generators run to meet demand at least cost. Frequency constrained versions usually lean on first order simplifications or single metrics like the frequency low point. Their model writes time domain frequency response with second order dynamics for accuracy.
Bernstein polynomial approximations turn the differential model into a tractable mixed integer program. A new constraint limits how long frequency stays below a critical threshold, guarding relays and equipment. Data driven links tie the below threshold area to excursion duration. Validation on a Spanish island grid shows stronger frequency security at moderate cost. Low inertia small systems stand to benefit most.
Catalog record
| CREATOR | Mohammad Rajabdorri, Bo Zhou, Lukas Sigrist, Enrique Lobato |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 43 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-02136
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Implementing General-Order Frequency Dynamic Response Model and Frequency Excursion Duration Criterion in Unit Commitment Problem Mohammad Rajabdorri, Bo Zhou, Lukas Sigrist, Enrique Lobato Abstract This paper introduces a novel approach for incorporating frequency dynamics into the unit commitment (UC) problem through a general-order differential equation model, solved using Bernstein polynomial approximation. Traditional frequency-constrained UC (FCUC) models typically rely on simplified first-order assumptions or scalar frequency metrics, such as frequency nadir, to indirectly enforce dynamic behavior. In contrast, our formulation explicitly models time-domain frequency response using second-order dynamics, enabling a more accurate and flexible representation of generator behavior. The resulting differential equations are approximated with high fidelity using Bernstein polynomials, leading to a mixed-integer linear programming (MILP) formulation that remains computationally tractable for small-scale power systems. Additionally, we introduce a new constraint based on the duration of frequency excursions below a critical threshold, motivated by practical concerns such as relay operation and equipment protection. A data-driven method is employed to relate the area under this threshold—computed as the integral of the Bernstein approximation—to the duration of frequency deviation. The proposed framework is validated using real-world data from an island system in Spain, demonstrating enhanced frequency security with a moderate increase in operational cost. These results suggest the method’s strong potential for application in low-inertia, small-scale power systems. I Introduction I-A Background and Motivation Maintaining power system stability is becoming increasingly challenging as traditional synchronous generators are progressively replaced by non-synchronous renewable energy sources. This challenge is especially critical in smaller power systems, where limited inertia significantly increases the risk of instability. For instance, in July 2020, the Spanish island of Tenerife experienced a complete blackout triggered by the shutdown of a power plant. The initial outage led to cascading failures and severe frequency deviations. Although the under frequency load shedding (UFLS) scheme was activated, demand reduction was not sufficient, ultimately resulting in a total blackout that lasted approximately seven hours [ 1 ] . To prevent such incidents in the operational stage, several preventive approaches have been proposed in the literature by integrating frequency-related constraints into scheduling problems, such as unit commitment. These methods aim to capture the frequency dynamics resulting from specific outage scenarios in order to avoid unacceptable frequency deviations [ 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 ] . Unlike traditional power system scheduling problems, typically solved with an hourly resolution, frequency response unfolds on a much faster timescale, often reaching critical thresholds within seconds. Moreover, these two domains differ significantly in mathematical formulation: scheduling problems are usually NP-hard and are often modeled as mixed integer linear programming (MILP) , whereas frequency dynamics are described using continuous-time ordinary differential equationss (ODEs) . These differences present significant challenges when attempting to incorporate frequency dynamics into conventional scheduling frameworks. The most common approach to incorporating frequency dynamics into power system scheduling is to extract key frequency metrics from simplified dynamic models and embed them into the optimization formulation [ 2 , 3 , 4 , 5 ] . Typical metrics include rate of change of frequency (RoCoF) , frequency nadir, and quasi-steady-state frequency. However, the effectiveness of these metrics is subject to debate. In practice, most equipment is sensitive not only to the magnitude of frequency deviations but also to the duration for which the frequency remains below a critical threshold [ 13 ] . Protective relays, for instance, are typically configured to respond based on the duration of such deviations according to their timeout settings [ 14 ] . Both analytical and data-driven techniques have been widely explored in the literature to estimate these frequency metrics. Analytical methods often rely on simplifying the well-known swing equation so that the aforementioned frequency metrics become mathematically tractable and can be directly incorporated into MILP -based scheduling problems. Among these metrics, the frequency nadir is particularly challenging to compute accurately. A common simplification assumes that generators respond linearly, delivering their reserves within a fixed time frame—typically set to 10 seconds. However, this assumption is questionable, as generator dynamics are more accurately described by second- or even third-order differential equations, depending on the technology and associated time constants [ 15 ] . To address this complexity and more accurately capture the reserve response of generators, data-driven methods have been proposed [ 7 , 8 , 9 , 10 , 11 , 12 ] . These approaches utilize detailed frequency response simulations to generate labeled datasets, from which machine learning models are trained to estimate frequency metrics. A key advantage of data-driven techniques lies in the accuracy of the labels, as the frequency metrics are directly derived from high-fidelity simulation results. However, a potential limitation is the inevitable loss of precision introduced by the learning algorithm and the quality of the training dataset. Bernstein polynomials form a basis for the space of continuous functions over a finite interval [ 16 ] . They allow for the accurate approximation of solutions to ODE problems while maintaining relatively low computational complexity [ 17 ] . In the context of power system operation and scheduling, Bernstein polynomials are particularly attractive due to their compatibility with MILP formulations [ 18 , 12 ] . For example, [ 6 ] applies Bernstein polynomials to model frequency response dynamics within the unit commitment (UC) framework. These and related studies are briefly reviewed in the following section. I-B Literature Review Table I summarizes recent contributions in the context of frequency nadir modeling within unit commitment. Of particular relevance to our work are the techniques used to approximate or predict frequency nadir and the assumptions to do so. TABLE I: Summary of Frequency Nadir Modeling in the Literature Ref. Frequency Nadir Modeling Underlying Assumptions [ 2 ] Inner approximation combined with binary expansion. Derived from the swing equation assuming linear primary response over 10 seconds, including load damping effects. [ 3 ] Piecewise linear upper and lower bounds. Based on the swing equation with a 10-second linear primary response, capturing fast dynamics of frequency-supporting resources. [ 4 ] Second-order conic reformulation of the nadir term. Derived from the swing equation, neglecting load damping effects. [ 5 ] Piecewise linear representation. Incorporates linear primary frequency responses from batteries, thermal units, and wind turbines using the swing equation. [ 6 ] Inner approximation of differential equations via Bernstein polynomials. Based on the swing equation with first-order differential-algebraic modeling of generator responses. [ 7 ] Constraint generation using optimal classification trees. Labeled using simulated trajectories from the swing equation with linear primary response. [ 8 ] Deep neural network prediction embedded in MILP. Data generated via positive-sequence power balance, labeled using a first-order frequency model. [ 9 ] Conservative sparse neural network predictor. Dataset generated using boundary and Latin hypercube sampling, labeled via the swing equation with first-order dynamics. [ 10 ] Logistic regression classifier. Labels derived from simulated frequency response using a second-order model. [ 11 ] Logistic regression and support vector machine ensemble. Frequency nadir labeled from a synthetically generated dataset using a second-order dynamic model. [ 12 ] Logistic regression on Bernstein polynomial coefficients. Labels based on a second-order frequency response model approximated via Bernstein polynomials. I-C Gaps,