A Corrective Frequency-Constrained Unit Commitment with Data-driven Estimation of Optimal UFLS in Island Power Systems Journal: International Journal of Electrical Power & Energy Systems

Miad Sarvarizadeh, Lukas Sigrist, Almudena Rouco, Mohammad Rajabdorri, Enrique Lobato · 2025

Read A Corrective Frequency-Constrained Unit Commitment with Data-driven Estimation of Optimal UFLS in Island Power Systems Journal: International Journal of Electrical Power & Energy Systems by Miad Sarvarizadeh, Lukas Sigrist, Almudena Rouco, Mohammad Rajabdorri, 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-2507-05062.txt.

Introduction

Sarvarizadeh and co-authors plan island power grids with learned shedding rules in a 2025 paper. Islands must commit generators ahead while guarding grid frequency after faults. When faults hit, shedding some load is the last defense, but planning for it is costly. The team teaches a model to guess the best shed amount from the first swing of frequency.

That guess folds into the planning problem, so cost and shedding are weighed together. Tests on a Spanish island grid with varied daily demand showed lower running costs. Frequency quality held, and shedding did not grow. Changing the price set on shedding shifted the balance as expected. The authors offer the method for cheaper, safer island grids.

Catalog record

CREATORMiad Sarvarizadeh, Lukas Sigrist, Almudena Rouco, Mohammad Rajabdorri, Enrique Lobato
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE39 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-05062

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

A Corrective Frequency-Constrained Unit Commitment with Data-driven Estimation of Optimal UFLS in Island Power Systems Journal: International Journal of Electrical Power & Energy Systems Miad Sarvarizadeh, Lukas Sigrist, Almudena Rouco, Mohammad Rajabdorri, Enrique Lobato Abstract This paper presents a novel corrective frequency constrained unit commitment (FCUC) formulation for island power systems by implementing data-driven constraint learning to estimate the optimal under-frequency load shedding (UFLS) . The Tobit model is presented to estimate the optimal amount of UFLS using the initial rate of change of frequency. The proposed formulation enables co-optimizing operation costs and UFLS . The aim is to account for optimal UFLS occurrences during operation planning, without increasing them. This would potentially reduce system operation costs by relaxing the reserve requirement constraint. The performance of the proposed formulation has been analyzed for a Spanish island power system through various simulations. Different daily demand profiles are analyzed to demonstrate the effectiveness of the proposed formulation. Additionally, a sensitivity analysis is conducted to demonstrate the effects of changing the cost associated with UFLS . The corrective FCUC is shown to be capable of reducing system operation costs without jeopardizing the quality of the frequency response in terms of UFLS occurrence. † † highlights: A novel corrective frequency-constrained unit commitment formulation is presented that considers and co-optimizes optimal under-frequency load-shedding amounts with system costs such that the system is secure after the outage of each generation unit, and operation costs are reduced by relaxing the reserve requirement. A data-driven constraint learning approach is proposed and implemented to estimate the amount of optimal under-frequency load shedding using the Tobit model. The estimating model is linearized and presented in mixed integer linear programming form to facilitate its addition to the standard unit commitment problem. The proposed formulation is applied to a real Spanish island power system. The effectiveness and capability of the formulation are analyzed through extensive simulations and sensitivity analyses. 1 Introduction The high penetration of renewable energy sources (RESs) in power systems has resulted in reduced inertia, as RESs do not inherently provide inertia. This issue is particularly critical in island power systems (IPSs) , where inertia is already limited due to the small number of available generating units [ 1 ] . The reduced inertia in IPSs lowers the overall frequency control capability, making the system more vulnerable to rapid frequency declines following the loss of a single generating unit. It is critical to address this issue in the scheduling problem of IPSs to avoid large frequency deviations. In most IPSs , the scheduling problem is centrally managed. The system operator sequentially performs unit commitment (UC) and economic dispatch (ED) over different time horizons. Optimal scheduling is typically constrained by security of supply requirements. A common security constraint is ensuring a sufficient amount of spinning reserve, which is activated in response to disturbances that cause an active power imbalance. Currently, the spinning reserve requirement follows a static criterion such that the amount of spinning reserve is sufficient to cover the largest possible contingency in the IPSs , following an N-1 security criterion. A known limitation of the static spinning reserve criterion is that it does not necessarily guarantee its timely activation [ 2 ] . After a disturbance, the spinning reserve is activated in real-time through primary and secondary frequency controls, where the primary frequency control restricts frequency variations and the secondary frequency control brings frequency back to its nominal value. However, in low-inertia systems, primary frequency control cannot always fully activate the spinning reserve in time to effectively contain frequency deviations. As a result, under-frequency load shedding (UFLS) is triggered to arrest frequency decay even under moderate disturbances in IPSs . This is mainly because the turbine-governor systems inherently exhibit a technology-dependent delay in their dynamic response to increase power. The dynamic nature of the problem of spinning reserve activation within the scheduling problem has already been addressed in the literature under the framework of frequency constrained unit commitment (FCUC) . This includes adding limitations on the post-contingency frequency characteristics such as rate of change of frequency (RoCoF) , frequency nadir, and quasi-steady-state frequency [ 3 , 4 , 5 , 6 , 7 , 8 ] . The frequency nadir was calculated in [ 3 ] under the assumption of a constant ramp rate in the overall mechanical power response, which is then used to formulate the corresponding constraints. The piecewise linearization technique is used to linearize the nonlinear function of frequency nadir in [ 4 ] , forming a FCUC formulation. The frequency response model for a multi-machine system is used to achieve higher accuracy. A mathematical model was proposed in [ 5 ] to represent changes in dispatchable units, assuming a linear transition rather than the typical staircase profile used in standard UC . The reserve constraint is then modified to ensure the feasibility of frequency regulation in the system. Separable programming is used in [ 6 ] to linearize and add the frequency nadir constraint to the UC problem. In [ 7 ] , the speed of response from different sources is modeled through distinct ramp rates. The resulting frequency nadir constraint is then linearized and added to the UC formulation. Similar to the previous reference, [ 8 ] proposes constraints for frequency nadir, RoCoF , and quasi-steady-state frequency addressing diverse frequency support resources and then linearizes the frequency nadir constraint using the linearized-frequency method. Data-driven constraint learning is proposed as an alternative to the analytical modeling of constraints. Constraint learning is the process of learning a function using data that relates the outcomes and decisions, to represent a constraint that is not clearly known or is highly nonlinear [ 9 ] . The learned function can then be used to generate predictions for a new observation. This method enhances the process of data-driven decision-making by combining machine learning techniques with optimization. Constraint learning has been used in various problems in the power systems field. In [ 10 ] , an ensemble learning method is used to describe the security boundary of the system. A security-constrained AC optimal power flow is presented in [ 11 ] that ensures small-signal stability and N-1 security, utilizing decision trees. [ 12 ] proposes a neural network method to obtain stability and security constraints in optimal power flow problems. While constraint learning has been applied to various power system studies, its role in frequency stability modeling is particularly relevant [ 13 , 14 , 15 , 16 ] . Logistic regression and support vector machines were proposed in [ 13 ] to predict the frequency nadir, with the resulting linear constraints incorporated into the UC problem to represent these models. An optimal classifier tree is proposed in [ 14 ] to form a FCUC that prevents frequency violations corresponding to the activation thresholds of the first UFLS relays. In [ 15 ] , a conservative sparse neural network method is given that approximates frequency nadir constraints and represents them in mixed integer linear programming (MILP) form. An extreme learning machine-based method is utilized in [ 16 ] to approximate the nonlinear frequency nadir constraint by a set of linear constraints. The above-mentioned literature is entirely focused on preventive measures to

Read the complete text