Source-linked AI summary
Controlled Tripping of Overheated Lines Mitigates Power Outages
René Pfitzner, Konstantin Turitsyn, Michael Chertkov
TL;DR
Fast blackout cascades can produce substantial damage, and the paper examines whether choosing the order of overloaded-line trips can reduce it. Using a quasi-static cascade model and state-based tripping heuristics on the Polish grid, the authors find that tripping order strongly affects outages and that controlled tripping can mitigate damage.
Problem
The paper investigates whether selecting the order of overloaded-line trippings can reduce damage from fast cascades, whose final outage varies substantially with the tripping path.
Source: §I · pp. 1–2Method
Results
Simple heuristics can mitigate severe cascades, with the heuristic that trips the least-loaded overloaded line performing best in most, but not all, cases.
Source: §I · p. 2Takeaways & Limitations
Controlled tripping can reduce cascade damage compared with waiting for natural tripping, but requires fast, reliable communication for state estimation, action selection, and execution.
Source: §I · p. 1The analysis focuses solely on tripping already overloaded lines, not other possible control actions such as generator dispatch or load shedding.
Source: §VI · p. 6
Abstract
from arXiv · showhide
We study the evolution of fast blackout cascades in the model of the Polish (transmission) power grid (2700 nodes and 3504 transmission lines). The cascade is initiated by a sufficiently severe initial contingency tripping. It propagates via sequential trippings of many more overheated lines, islanding loads and generators and eventually arriving at a fixed point with the surviving part of the system being power-flow-balanced and the rest of the system being outaged. Utilizing an improved form of the quasi-static model for cascade propagation introduced in our earlier study (Statistical Classification of Cascading Failures in Power Grids, IEEE PES GM 2011), we analyze how the severity of the cascade depends on the order of tripping overheated lines. Our main observation is that the order of tripping has a tremendous effect on the size of the resulting outage. Finding the "best" tripping, defined as causing the least damage, constitutes a difficult dynamical optimization problem, whose solution is most likely computationally infeasible. Instead, here we study performance of a number of natural heuristics, resolving the next switching decision based on the current state of the grid. Overall, we conclude that controlled intentional tripping is advantageous in the situation of a fast developing extreme emergency, as it provides significant mitigation of the resulting damage.
I. INTRODUCTION
The paper asks whether selecting the order of overloaded-line trippings can limit fast cascade damage, rather than waiting for natural trippings. On the Polish grid model, it finds that outcomes vary sharply by tripping path and that simple state-based heuristics can mitigate severe cascades, though reliable communications are needed for emergency control.
I. INTRODUCTION
- Fast emergency control requires on-the-fly state estimation and action selection, flawless execution, and fast, reliable communications.The paper focuses on tripping already overloaded lines, rather than other possible control actions such as generator dispatch or load shedding.Source: §I · p. 1
- The study uses the Polish transmission-grid model and compares four memoryless heuristics that choose the next overloaded line from the grid’s current state.Cascade damage is measured by the fraction of demand not served and by the number of steps until the cascade stops.Source: §I · p. 2
- For the same initial contingency, random tripping paths produce a broad range of outcomes, from almost no outage to a very significant loss of grid service.This variation motivates choosing line-tripping actions rather than waiting for probabilistically natural trippings.Source: §I · p. 2
- Some simple graph-local heuristics mitigate severe cascades, and the least-loaded overloaded-line heuristic performs best in most, but not all, tested cases.The authors relate its performance to a simplified hierarchical interpretation of the grid.Source: §I · p. 2
II. OUR MODEL
The model represents a cascade as sequential line trips and resulting islanding, load shedding, and generation adjustment. It assumes fast generation redispatch is unavailable, and iterates until the surviving grid reaches a feasible steady state.
II. OUR MODEL
- Fast generation redispatch is assumed unavailable, so the model relies on primary droop control and emergency load shedding during the emergency timescale.Generation redispatch is typically performed every 15 minutes to an hour, according to the model assumptions.Source: §II · p. 2
- The cascade starts from an Optimal Power Flow solution with one or more initial line failures, then checks islanding and uses droop control to match generation and demand.Source: §II · p. 2
- After an initial contingency, the model repeatedly resolves power flows, trips one overloaded line, checks for islands, and applies load shedding or island outages as needed.A cascade may begin when a small number of coincidental failures overload other lines; without operator action, an overloaded line is likely to trip within minutes.Source: §II · pp. 2–3
- The cascade algorithm repeats until thermal constraints are resolved and a steady feasible solution is achieved on the surviving system.The single-line-at-a-time choice reflects uncorrelated natural trips and is also described as safer for control actions than simultaneous trips.Source: §II · p. 3
A. DC power flow
The model uses power-flow equations to map nodal injections and system parameters to node phases and line flows. It chooses the simpler DC approximation to focus on tripping sequences and network effects, while noting that the framework can use more accurate AC power flow.
A. DC power flow
- A general power-flow solver takes nodal power injections and consumptions plus system parameters, and returns node voltages and phases and transmission-line powers.Source: A. DC power flow · p. 3
- The study uses a simpler DC solver, which ignores voltage and resistive-line variations and assumes small phase differences across lines.The authors say this approximation lightens the algorithm and focuses analysis on tripping sequences and network effects; the framework can also use AC power flow.Source: A. DC power flow · p. 3
- In the DC equations, x, g, d, θ, and p denote line inductances, generator injections, load demands, node phases, and line flows, respectively.Source: A. DC power flow · p. 3
B. Optimal Power Flow
The base solution is obtained by solving a DC optimal power-flow problem using the initial load, generator cost functions, and generation and line-capacity constraints. Its resulting flows, generation, and phases serve as the cascade algorithm’s reference state.
B. Optimal Power Flow
- The base solution solves the standard DC optimal power-flow problem for generator dispatch given initial load, generator cost functions, and generation and line-capacity constraints.Source: B. Optimal Power Flow · p. 3
- The study uses MATPOWER and the generator cost functions supplied with the Polish grid model to compute the base solution.Source: B. Optimal Power Flow · p. 3
- The resulting p0, g0, and θ0 form the base reference solution for the cascading algorithm.Source: B. Optimal Power Flow · p. 3
C. Identify islands
The algorithm detects islands as the grid splits and handles cascade computations independently within each island.
C. Identify islands
- After line tripping forms islands, the algorithm checks for them and performs subsequent cascade computations independently for each island.It uses a depth-first-search-based algorithm to detect grid components.Source: C. Identify islands · p. 3
D. Droop Control and Load Shedding
When tripping disconnects generators or loads or forms islands, the model redistributes generation through local droop control and applies load shedding if an island cannot balance.
D. Droop Control and Load Shedding
- Line tripping can disconnect loads or generators or split the grid into islands, requiring fast automatic redistribution of generation.The model uses droop control for this redistribution.Source: D. Droop Control and Load Shedding · pp. 3–4
- Droop control instantaneously redistributes generation among generators in a newly formed island according to their current generation and the island's total generation and demand.If a generator reaches its capacity, it is held at that level and excluded from further droop redistribution.Source: D. Droop Control and Load Shedding · p. 4
- An island persists if its demand can be matched by generation; otherwise, the model first sheds 10% of its loads and shuts the island down only if balance remains impossible.The authors compare this load-shedding scheme with an alternative that immediately switches off an overloaded island.Source: D. Droop Control and Load Shedding · p. 4
III. RANDOM (NATURAL) TRIPPING
The authors sample random natural tripping paths for selected severe initial contingencies and find that outcomes vary widely, making the order of tripping consequential.
III. RANDOM (NATURAL) TRIPPING
- Random tripping samples each overheated line uniformly at each step, and the authors repeat the resulting paths for four selected initial contingencies.The cases are single-line trips of lines 44, 2832, and 102, and a simultaneous trip of lines 3 and 29.Source: §III · p. 4
- The selected single-line contingencies come from pre-sampled severe cases among the top 1% most-loaded lines, while the two-line case is likewise selected from severe pre-sampled combinations.Most pre-sampled cases did not produce notable outages.Source: §III · p. 4
- Across all four initial contingencies, random tripping produces broad outage-size distributions, with both little or no damage and severe outages possible.Three cases have a significant number of zero- or small-damage outcomes; even the line-44 case has some.Source: §III · p. 4
- The four selected contingencies have different average outage sizes despite causing approximately the same initial power loss.The line-44 and simultaneous line-3-and-29 cases have significantly larger average damage than the other two.Source: §III · p. 4
- Cascade length and resulting blackout size are strongly correlated in the plotted cases.Cascade length is measured by the number of lines tripped, and blackout size by the fraction of load not served.Source: §III · p. 4
IV. EFFECT OF LOAD-SHEDDING
Excluding load shedding produces no significant qualitative change in the outage distribution, though low-outage scenarios perform slightly better without it.
IV. EFFECT OF LOAD-SHEDDING
- Excluding load shedding causes no significant qualitative change in the outage distribution, while low-outage scenarios perform slightly better without it.The comparison switches overloaded islands off immediately when load shedding is excluded.Source: §IV · p. 5
V. CONTROL OF THE TRIPPING PATH
The authors test four memoryless, state-based rules for choosing which overloaded line to trip, because outage size is highly sensitive to tripping order. The least-loaded rule performs best in most tested cases, but no heuristic is universally reliable.
V. CONTROL OF THE TRIPPING PATH
- The four memoryless heuristics choose the next overloaded line by minimum or maximum current flow or relative overload.A1 and A3 minimize current flow and relative overload, respectively; A2 and A4 maximize those measures.Source: §V · p. 5
- A1’s hierarchical rationale is that tripping a low-flow line may shut down a small sub-tree, shed little demand, and reduce stress on higher-level lines.The authors note that the Polish grid is not cleanly hierarchical, though its size and local structure suggest some tree-like character.Source: §V · p. 6
- The tested decisions rely on fast communication of overloaded-line status to a central location and transmission of a tripping signal back to the grid.The reported histograms show outage-size distributions for samples initiated by different line contingencies; insets include the largest outages and average cascade length.Source: Figure 3 · p. 5, Figure 4 · p. 5, Figure 5 · p. 5
- Across comparisons with other schemes and uniform random trippings in Figs. 3–6, A1 was a clear-cut winner in all but the third Fig. 5 example, while tests were inconclusive about a universally reliable heuristic.Uniform tripping yielded no or almost no damage for a significant O(1) number of samples in Figs. 4–6.Source: §V · p. 6
VI. SUMMARY
The study finds that controlled tripping of already overloaded lines may mitigate cascade damage, while a universally optimal tripping sequence is computationally hard. Its analysis is limited to line tripping in a Polish-grid model using a microscopic DC power-flow approximation.
VI. SUMMARY
- Controlled tripping of overloaded lines may reduce damage, but finding a universally optimal sequence is computationally hard.The study tests graph-local heuristics based on the grid’s current state; the best-performing strategies in most cases trip the least-overloaded line first.Source: §VI · p. 6
- The analysis considers only already-overloaded-line tripping on a real-world grid structure using a microscopic DC power-flow approximation.The authors identify generalization to AC and hybrid schemes combining line tripping with other actions as future work.Source: §VI · pp. 6–7