Inferensys

Glossary

Under-Frequency Load Shedding (UFLS)

An automatic, last-resort protection scheme that disconnects predetermined blocks of customer load in a progressive manner to arrest a severe and rapid decline in system frequency and prevent a total blackout.
Developer building agentic RAG system, retrieval pipeline diagram on laptop, technical workspace with notes.
LAST-RESORT PROTECTION SCHEME

What is Under-Frequency Load Shedding (UFLS)?

An automatic, emergency control scheme designed to prevent a catastrophic power system collapse by rapidly disconnecting predetermined blocks of customer load when system frequency falls below critical thresholds.

Under-Frequency Load Shedding (UFLS) is an automatic, decentralized protection scheme that arrests a severe and rapid decline in system frequency by progressively disconnecting predetermined blocks of customer load. It functions as the final safety net when primary frequency response and contingency reserves are insufficient to rebalance generation and load following a major disturbance, such as the sudden loss of a large generating unit.

UFLS relays, deployed at distribution substations, operate autonomously based on locally measured frequency. They are configured with discrete frequency setpoints—typically between 59.3 Hz and 58.5 Hz—and intentional time delays. When frequency drops below a setpoint for the specified duration, the relay trips a feeder breaker, shedding a calculated percentage of the balancing authority's total load to prevent a cascading, interconnection-wide blackout.

LAST-RESORT PROTECTION

Core Characteristics of UFLS

Under-Frequency Load Shedding (UFLS) is not a control algorithm but a pre-programmed, automatic defense mechanism. It sacrifices predetermined blocks of load to prevent a catastrophic system-wide collapse when frequency plummets.

01

The Last Line of Defense

UFLS is the final safety net after Primary Frequency Response (governor action) and Automatic Generation Control (AGC) have failed to arrest a frequency decline. It is a remedial action scheme designed for extreme contingencies where generation loss vastly exceeds the available spinning reserve. Without UFLS, a cascading blackout is mathematically certain.

02

Progressive, Frequency-Based Tripping

Load is shed in discrete blocks (typically 5-10% of total load per step) at predefined frequency thresholds. A common scheme might trip:

  • Block 1: 59.3 Hz (lightest penalty, fastest action)
  • Block 2: 59.0 Hz
  • Block 3: 58.7 Hz
  • Block 4: 58.4 Hz (heaviest penalty, last resort) Each block includes an intentional time delay (usually 6-30 cycles) to override transient dips.
03

Arresting the Frequency Nadir

The primary objective is to halt the frequency decline at a safe nadir (minimum point) above the critical threshold for thermal turbine blade resonance (typically around 57.5 Hz). By rapidly balancing generation and load, UFLS prevents the under-frequency protection relays on generating units from tripping them offline, which would accelerate the collapse.

04

Mandated by NERC Reliability Standards

In North America, UFLS programs are mandatory per NERC Reliability Standard PRC-006. Regional entities define the exact parameters, but the standard requires:

  • Shedding a minimum total percentage of peak load.
  • Tripping blocks within defined frequency and time delay ranges.
  • Preventing inadvertent restoration of shed load until frequency recovers. Compliance is non-negotiable for Balancing Authorities and Transmission Owners.
05

Dynamic vs. Static Shedding Logic

Traditional UFLS is static—the amount of load shed is fixed regardless of the magnitude of the disturbance. Advanced Adaptive UFLS schemes use real-time Rate of Change of Frequency (ROCOF) measurements to estimate the total generation deficit. This allows the system to shed the exact amount of load needed in fewer steps, minimizing customer impact while maximizing grid survival probability.

06

Under-Voltage Load Shedding (UVLS) Coordination

UFLS must be carefully coordinated with Under-Voltage Load Shedding (UVLS) schemes. A severe disturbance can cause both frequency and voltage to collapse simultaneously. Uncoordinated tripping can exacerbate the imbalance. Modern Remedial Action Schemes (RAS) integrate both UFLS and UVLS logic to ensure the correct load is shed based on the dominant grid instability mode.

UNDER-FREQUENCY LOAD SHEDDING

Frequently Asked Questions

Critical questions about the automatic, last-resort protection scheme that prevents total system collapse during severe generation-load imbalances.

Under-Frequency Load Shedding (UFLS) is an automatic, last-resort protection scheme that disconnects predetermined blocks of customer load in a progressive manner to arrest a severe and rapid decline in system frequency and prevent a total blackout. It operates as a decentralized, relay-based defense mechanism independent of the central Automatic Generation Control (AGC) system.

When a sudden loss of generation occurs, the immediate power imbalance causes system frequency to decay at a rate proportional to the magnitude of the deficit. UFLS relays, installed at distribution substations, continuously monitor local frequency. When frequency drops below a predefined threshold—typically between 59.3 Hz and 58.5 Hz in a 60 Hz system—the relays trip their assigned feeder breakers after a short intentional time delay, shedding a calculated percentage of total system load.

The scheme is designed in discrete stages or blocks, with each stage shedding an incremental amount of load at progressively lower frequency setpoints. For example, Stage 1 might shed 10% of load at 59.3 Hz, Stage 2 another 10% at 59.0 Hz, and Stage 3 an additional 10% at 58.7 Hz. This graduated approach ensures that only the minimum necessary load is interrupted to stabilize the grid, avoiding over-shedding that could cause an over-frequency excursion.

Prasad Kumkar

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.