Definition
Ratio of load-driven equipment overloads or failures that occurred within areas the AI load forecast identified as having adequate capacity: measuring whether the forecast's spatial granularity matches the granularity at which real overloads occur.
The Measurement Gap
Traditional load forecasting operates at the substation or planning area level. Engineers have used this resolution for decades because that was the level at which capacity decisions (new substations, feeder upgrades, transformer banks) were made. That resolution worked when load grew slowly and uniformly. It does not work when load growth is lumpy, clustered, and technology-driven.
The specific failure point is spatial mismatch. An AI load forecast may predict that Substation A's service territory will grow from 85 MW peak to 92 MW peak over the next three years, and that forecast may be exactly right. But within that territory, a 20-home neighborhood with six early EV adopters on one residential transformer may already be seeing peak loads that exceed the transformer's nameplate rating on summer evenings. The aggregate forecast is accurate. The transformer fails. SAIDI records the outage. Nobody connects it to the forecast's spatial resolution gap.
LFSRA enables the specific decision of whether to increase the load forecast's spatial granularity from substation to feeder or transformer level, or to supplement the forecast with bottom-up EV and DER adoption data at the circuit level. That decision requires knowing how frequently the current forecast fails to predict where load-driven failures actually occur.
Without this measurement, planning engineers invest in substation-level forecast accuracy while distribution transformers fail in patterns the forecast cannot see. The infrastructure investment is misaligned because the planning tool's resolution doesn't match the failure granularity.
AMI interval data at the transformer secondary and advanced GIS integration make circuit-level and near-transformer load estimation feasible in ways that were not possible with monthly billing data and paper maps.
Formula
LFSRA = Overloads_in_forecast_adequate_areas / Total_load_driven_overloads
Variables:
- Overloads_in_forecast_adequate_areas: count of load-driven thermal events, distribution transformer failures, or feeder overloads in geographic areas where the AI load forecast showed capacity headroom (load below rated capacity threshold) at the time of the planning cycle that preceded the event (count). Source: OMS outage records matched to GIS forecast zones
- Total_load_driven_overloads: all load-driven equipment distress events (thermal overloads, transformer failures attributed to load, feeder capacity limit exceedances) in the measurement period (count). Source: OMS, CMMS failure records with cause codes
- Capacity headroom threshold: typically defined as actual or forecast peak load below 85–90% of equipment nameplate rating. Source: distribution planning model
Calculation Frequency: Annual (requires 2–3 year accumulation for utilities with low annual overload counts); semi-annual in high-growth service territories
Output: Percentage, 0% to 100%. Below 20% means overloads are concentrated where the forecast predicted stress: the forecast is spatially useful. Above 50% means more than half of overloads are occurring as surprises in areas the forecast said were fine. EV-driven overloads should be tracked as a sub-category when data permits.
Proposed formula, pending validation pilots.
At a Glance
Decision Supported
Is the load forecast accurate at the level where failures actually happen, or is it accurate at the substation level while transformer-level overloads go unpredicted?