⚡ electrical engineering

OPF-Based Optimal Power System Network Restoration Considering Frequency Dynamics

This paper proposes an optimal power system restoration framework that integrates frequency dynamics to address stability challenges in low-inertia islands, demonstrating through the IEEE 9-Bus model that static optimization alone is insufficient as it can lead to sequences violating dynamic constraints.

Original authors: Dawn Virginillo, Asja Derviškadić, Mario Paolone

Published 2026-04-17

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Dawn Virginillo, Asja Derviškadić, Mario Paolone

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 massive, intricate city of roads. When a massive blackout hits, it's like a total traffic jam where all the lights go out, and the roads are empty. Power System Restoration (PSR) is the plan to get the city moving again.

Usually, engineers have a "static" plan for this. They look at the map and say, "Okay, we have one power plant (a generator) and three neighborhoods (loads). Let's turn on the power plant, then connect Neighborhood A, then B, then C." They check if the math adds up: Does the power plant have enough energy to light up all three neighborhoods? If the numbers say "yes," the plan is approved.

But here's the problem: This paper argues that the static plan is like planning a road trip without checking if the car can actually handle the hills.

The Core Problem: The "Too Fast, Too Heavy" Trap

In the real world, electricity isn't just a number; it has momentum (inertia). Think of the power grid like a giant spinning flywheel.

The Static Plan: It assumes the flywheel is infinitely heavy and unshakeable. It says, "We can hook up 100 houses to 1 generator instantly!"
The Reality: If you hook up 100 houses to 1 generator too quickly, the generator is like a cyclist trying to pedal a heavy truck up a steep hill. The cyclist (the generator) gets overwhelmed, slows down, and eventually stops (the frequency drops). If it drops too low, the safety systems kick in and shut the generator down to save it. The whole plan fails, and the lights stay off.

This paper introduces a new way to plan: The Dynamic Plan.

The Solution: "DynOPF-R" (The Smart Navigator)

The authors created a new mathematical tool called DynOPF-R. Instead of just looking at the map (static power flow), this tool simulates the physics of the ride.

Here is how it works, using simple analogies:

1. The "Baby Steps" Approach

In the old static plan, the computer might say, "Turn on Generator 1, then immediately flip switches for 5 neighborhoods."
In the new Dynamic Plan, the computer acts like a cautious driver. It says, "Okay, we can turn on Generator 1. But before we add Neighborhood A, we need to wait a few seconds to make sure the engine (frequency) doesn't stall. Then, we add Neighborhood A. Wait. Check the engine speed. Okay, now we can add Neighborhood B."

It calculates the frequency (the speed of the grid's "heartbeat") at every single second. If adding a load makes the heartbeat skip a beat, the plan automatically changes to a safer route.

2. The "Tightrope Walker"

Imagine the grid frequency is a tightrope.

Static View: The walker just needs to get from Point A to Point B.
Dynamic View: The walker must balance perfectly. If they step too heavy (add too much load too fast), they fall off the rope (frequency violation).
The new algorithm finds the path that gets the most people across the rope without anyone falling. It might take a slightly longer route or wait longer between steps, but it guarantees the walker stays on the rope.

3. The "Two-Island" Dance

Sometimes, during a blackout, the grid splits into two separate islands (like two separate towns that lost connection). To fix this, you have to reconnect them.
The paper shows that their new tool can handle this "dance." It ensures that when two islands try to merge, they are spinning at the exact same speed. If they try to merge while one is spinning fast and the other slow, it's like two gears grinding together—it breaks the machine. The new tool calculates the perfect moment to snap them together.

Why Does This Matter?

The authors tested this on a standard model (the IEEE 9-Bus system, which is like a small test city).

The Old Way: The computer found a "perfect" plan that turned on 5 neighborhoods using just 1 generator. It looked great on paper.
The New Way: When they ran the "physics simulation" on that old plan, the frequency crashed. The generator would have tripped offline, and the restoration would have failed.
The Result: The new tool found a different plan. It turned on fewer neighborhoods first, waited for the generator to stabilize, and then added more. It took a bit more time, but it actually works.

The Big Takeaway

In the past, engineers often used "rules of thumb" (heuristics) or checked the physics after they made a plan. This paper says: "Don't check the physics later; build the physics into the plan from the start."

By using advanced math (Mixed-Integer Programming) to simulate the real-time "heartbeat" of the grid, this new method ensures that when we try to bring the lights back on, we don't accidentally blow the fuse again. It turns restoration from a game of "guess and hope" into a precise, physics-based science.

In short: It's the difference between a plan that says "Drive fast to get there" and a plan that says "Drive fast, but watch the speedometer so you don't crash."

1. Problem Statement

Power System Restoration (PSR) is the process of re-energizing a grid after a total or partial blackout. Traditional PSR planning often relies on heuristics or static Optimal Power Flow (OPF) models that assume steady-state conditions. However, modern grids face increased complexity due to Distributed Energy Resources (DERs) and reduced system inertia (low-inertia islands).

The core problem addressed is that static optimization models fail to account for transient frequency dynamics. A switching sequence that appears optimal under static constraints (e.g., maximizing load pickup quickly) may cause severe frequency deviations (nadir violations) during the transient phase, leading to generator trips and restoration failure. The paper aims to formulate and solve an Optimal PSR problem that explicitly integrates frequency dynamics to ensure both static feasibility and dynamic stability.

2. Methodology

The authors propose a Mixed-Integer Programming (MIP) framework that integrates power flow constraints with dynamic system equations. The methodology is divided into two main formulations:

A. Static Formulation (OPF-R)

Basis: Uses a DC Optimal Power Flow (DC OPF) approximation, neglecting voltage and reactive power to maintain linearity and computational tractability.
Objective: Maximize the total energy served over a fixed restoration horizon.
Constraints:
- Power balance (linearized).
- Switching logic: Components (loads, lines, transformers, generators) can only be energized if connected to an already live bus.
- One component can be switched per timestep.
- Generator limits and line flow limits.
Validation: The authors validate the global optimality of the switching constraints in this static version using a brute-force depth-first tree search on a reduced IEEE 9-Bus model, confirming that the MIP solver finds the global optimum.

B. Dynamic Formulation (DynOPF-R)

This is the paper's primary contribution. It extends the static model to include frequency dynamics using a single-stage optimization approach.

Dynamic Model:
- Governor & Turbine: Modeled using first-order differential equations representing the IEEEG3 governor model, including transient and steady-state droop control.
- Swing Equation: Models the mechanical dynamics of synchronous machines ( $M\dot{\omega} = P_m - P_e - D\omega$ ).
- Integration: The differential equations are discretized using the Backward Euler method to maintain linear constraints within the MIP framework.
Coupling Mechanisms:
- Big-M Constraints: Used to handle the initialization of Non-Black Start Units (NBSUs). Specifically, they enforce that an NBSU's frequency must match the island frequency before it can be switched in (simulating synchrocheck).
- Power Balance: Modified to link electrical power output ( $P_e$ ) directly to the dynamic swing equation.
Objective Function: Modified to include a penalty term for frequency deviations ( $\beta \sum |\Delta \omega|$ ) alongside the energy served term ( $\alpha \sum S_d P_d$ ).
Time Discretization: Introduces a "dead time" parameter ( $\kappa$ ) between switching events to reduce the dimensionality of the problem and improve solver efficiency.

3. Key Contributions

Integrated Single-Stage Formulation: Unlike previous works that use multi-stage or iterative approaches (optimizing switching then checking dynamics), this paper proposes a single-stage MIP that simultaneously optimizes the switching sequence and generator setpoints while satisfying dynamic constraints.
Physics-Informed Dynamics: The model incorporates detailed governor dynamics (transient droop) and swing equations, allowing for validation against detailed Electromagnetic Transient (EMT) models.
Global Optimality Validation: The authors demonstrate a method to validate the global optimality of the switching constraints using tree-search methods on the static formulation.
Demonstration of Static vs. Dynamic Discrepancy: The paper provides concrete evidence that a switching sequence optimal for static constraints is often dynamically infeasible, causing frequency violations.

4. Results and Case Studies

The formulations were tested on the IEEE 9-Bus model with parameters inferred from a reference EMT model to ensure realistic dynamic behavior.

Static vs. Dynamic Comparison:
- Static (OPF-R): The optimal solution energized 5 loads at Bus 5 and Bus 6 using only one generator before adding a second generator.
- Dynamic (DynOPF-R): The optimal solution prioritized energizing loads at Bus 6 first, then adding a second generator (G2) before picking up the heavy loads at Bus 5.
- Outcome: When the static solution (1 generator + 5 loads) was simulated with dynamic constraints, it resulted in a frequency nadir below the safety limit (violating the 48 Hz lower bound), causing a hypothetical trip. The dynamic formulation successfully avoided this by adjusting the sequence.
Parameter Inference: The authors successfully inferred dynamic parameters (Inertia $M$ , Damping $D$ , Governor gains) from a reference EMT model, showing that the simplified DC OPF + dynamic equations can approximate EMT behavior with low Root Mean Squared Error (RMSE).
Multi-Island Scenario: A second case study involving two initially energized islands (G1 and G2) demonstrated the model's ability to handle island synchronization. The solver correctly identified the optimal time to close the tie-line (at $t=180$ s) to synchronize the islands before further load pickup.
Computational Performance:
- Static formulation solved in seconds.
- Dynamic formulation required significant time (e.g., 12.3 hours for the 9-Bus case) due to the combinatorial complexity and the large number of timesteps (3600 steps) required for accurate dynamic simulation.

5. Significance and Conclusion

Operational Relevance: The study highlights that relying on static heuristics for backbone construction in low-inertia systems is risky. The proposed DynOPF-R provides a rigorous mathematical tool to generate restoration plans that are guaranteed to respect frequency stability limits.
Future-Proofing: As grids transition to higher renewable penetration (lower inertia), the ability to dynamically optimize restoration sequences becomes critical.
Limitations & Future Work:
- The current model uses DC OPF, which neglects voltage and reactive power; future work aims to incorporate AC constraints.
- Computational time is currently high for large-scale systems; scalability improvements are needed.
- The impact of energy storage systems (BESS) on restoration dynamics is not yet included.

In summary, this paper establishes a foundational framework for dynamic-aware optimal power system restoration, proving that integrating frequency dynamics into the optimization loop is essential for preventing restoration failures in modern, low-inertia power grids.