Engineering Specification — Normative · Version 1.0 · July 8, 2026

The Coupling-Discrimination Specification

Separating Orchestration-Coupling from Environmental Correlation

Abstract. This specification defines the criteria by which a load is admitted into a behaviorally-coupled block for the Minimum-Inertia absorption test. It separates orchestration-coupling (loads moved by a shared control plane — a common-mode contingency) from environmental correlation (loads independently responding to a common signal — legitimate demand response). It mandates classification on four boundary-observable signatures — timing coherence, profile congruence, residual correlation after conditioning on environmental signals, and participation consistency — and defines the residual coupling coefficient ρ_coupling as the admitting invariant. It ensures the absorption test constrains only genuine common-mode risk and does not penalize independent flexibility.

Requirement language: RFC 2119 (SHALL / MUST / MUST NOT are normative).

1. Requirement — correlation alone is insufficient

Two loads may exhibit correlated power trajectories for physically distinct reasons: a shared control plane commanded them (orchestration-coupling — the contingency), or each independently responded to the same public signal such as price or system frequency (environmental correlation — the demand-response behavior the system relies upon). Both produce high aggregate correlation.

Normative. Admission of a load into a coupled block MUST NOT be based on aggregate correlation alone. A load MUST be admitted only when the joint signature condition of Section 2 is satisfied.

2. Normative discriminating signatures

2.1 Timing coherence

A shared command arrives within a narrow machine-speed window; independent responses disperse over each participant's own reaction time. The specification SHALL define a coincidence window Δt at the machine-speed scale. Onsets clustering within Δt are candidate-orchestrated; onsets dispersed beyond Δt are candidate-environmental.

2.2 Profile congruence

A command replicates one instruction, producing near-identical normalized ramp waveforms; independent responses vary in shape. The specification SHALL evaluate the shape correlation of normalized ramp waveforms, not merely aggregate correlation.

2.3 Residual correlation after environmental conditioning (normative control)

Before attributing correlation to coupling, the specification MUST regress each load's behavior against the observable common environmental signals — nodal price, system frequency, published demand-response dispatch, and ambient temperature. Only the correlation of the residuals — the co-movement remaining after those signals are removed — MUST be counted toward the coupling coefficient:

ρ_coupling = corr( residualx , residualj )
after conditioning on common environmental signals

A set of loads responding to a common price signal exhibits high raw correlation but near-zero residual correlation and MUST be cleared as environmental. A set sharing a scheduler exhibits correlation that survives conditioning and MUST be identified as orchestrated.

2.4 Participation consistency

Independent loads occasionally diverge; orchestrated loads execute with high completion. The specification SHALL evaluate response-completion consistency across repeated events; near-complete joint participation is a signature of common control, natural scatter a signature of independence.

3. Normative admission and the load-migration fraction

Normative. A load pair (and, extended, a cluster) SHALL be classified as orchestration-coupled only when all of the following hold jointly: timing coherence within Δt; profile congruence above the shape-correlation bound; residual coupling ρ_coupling above the coefficient threshold after environmental conditioning; and participation consistency above the completion bound. Loads not satisfying the joint condition MUST NOT be admitted to the coupled block.

The load-migration fraction — the input to ΔP_block in the Minimum-Inertia Specification — SHALL be the largest sum of mutually orchestration-coupled load, expressed as a fraction of interconnected load, admitted under this section. Environmentally-correlated load MUST be excluded from this fraction.

4. Party classification (normative outcome)

The specification resolves each participant into exactly one measured position:

Because position is established by observable behavior against a stated threshold, no party is admitted to a contingency obligation by mere correlation, and no party escapes one by disputing intent. The classification is auditable and reproducible: any party with access to the same boundary telemetry and the published environmental signals can independently recompute the coupling coefficient and verify the resulting classification. No internal or proprietary facility data is required.

5. Application to coordinated mechanisms

Any coordinated response that directs multiple participants to change consumption on a common trigger satisfies the orchestration signatures and SHALL be evaluated as a coupled block under the Minimum-Inertia Specification, irrespective of how the mechanism characterizes itself. Whether a coordinated action supports or threatens stability MUST be resolved by the absorption inequality, not by label.

Appendix A — Invariant reference

SymbolDefinitionVerification
ρ_couplingResidual coupling coefficient after conditioning on common environmental signals. The conditioning set comprises the publicly observable signals that could independently drive co-movement — nodal/locational price, system frequency, published demand-response or dispatch instructions, and ambient temperature — chosen because they are available to all loads simultaneously and are therefore the plausible sources of non-orchestrated correlation.Regression residual correlation
ΔtCoincidence window at machine-speed scaleRamp-onset dispersion analysis
migration fractionLargest mutually orchestration-coupled load fractionJoint-signature admission