Generator Coherency

The fundamental criterion for coherency is that generators exhibit synchronized rotor angle deviations following a disturbance. If the angular difference between two generators remains constant over time, they are considered coherent. This is formally expressed as:

• Δδᵢⱼ(t) = δᵢ(t) - δⱼ(t) ≈ constant
• The rate of change of the angular separation approaches zero
• Generators accelerate and decelerate in unison

This behavior arises because coherent machines share similar electrical proximity to the fault location and possess comparable inertia constants.

Within a coherent group, the inter-machine oscillations are negligible compared to the inter-group oscillations. The internal dynamics damp out rapidly, allowing the group to be treated as a single rigid mass.

• Intra-group modes have high frequency and high damping ratios
• Inter-group modes dominate the system response
• The slow coherency property ensures that internal connections are stiff relative to external tie-lines

This separation of timescales is the mathematical basis for Kron reduction and dynamic equivalencing.

Coherency is strongly influenced by the electrical distance between generators. Machines that are electrically close—separated by low-impedance paths—tend to swing together because they experience similar voltage depressions during faults.

• Short transmission lines with low reactance promote coherency
• Generators behind a common bus or substation are natural candidates
• Topological clustering in the admittance matrix reveals candidate groups

Graph-theoretic partitioning of the network based on edge weights (line admittances) often pre-identifies coherent clusters before dynamic simulation.

Generators with comparable inertia constants (H) and similar governor-turbine dynamics will exhibit matched frequency responses. The ratio of inertia to accelerating power determines the initial rate of change of speed.

• Hᵢ ≈ Hⱼ leads to similar RoCoF profiles
• Machines with identical droop characteristics share load proportionally
• The equivalent machine's inertia is the sum of individual inertias: H_eq = Σ Hᵢ

Mismatched inertia within a group causes internal swinging, violating the coherency assumption and requiring subgroup partitioning.

Coherency is not an absolute property of generators; it is a function of the disturbance location. A group that is coherent for a fault at Bus A may split into subgroups for a fault at Bus B.

• Coherency maps are generated for each critical contingency
• The accelerating power distribution changes with fault position
• Slow coherency is disturbance-independent; transient coherency is fault-specific

Practical dynamic equivalencing requires identifying groups that remain coherent across a credible contingency set, not just a single event.

Coherency identification is verified through numerical integration of the full nonlinear swing equations. The angular separation between candidate generators is monitored over a 2-5 second simulation window.

• RMS coherency index quantifies the average angular deviation
• Generators are clustered using hierarchical agglomerative algorithms on the distance matrix
• The silhouette score validates cluster quality

Machine learning classifiers, such as support vector machines and graph neural networks, are now used to predict coherency directly from pre-fault operating conditions without time-domain simulation.