In short. A study is only as good as its inputs. Every study should list the data used, where it came from, what was assumed and how it was checked. Validation against nameplates and measurements, and a sensitivity check on the weakest assumptions, are what make the result defensible.
Input checklist
| Group | Data |
|---|---|
| Network | Single-line diagram, voltage levels, utility fault level and X/R, operating modes |
| Sources | Generator ratings, reactances, excitation, governor data, UPS and BESS data |
| Transformers | kVA, impedance, vector group, tap, losses |
| Cables and busbars | Type, size, length, route, ratings |
| Loads | kW, power factor, starting method, duty, harmonic data |
| Protection | Device types, models and settings |
| Site | Altitude, temperature, soil resistivity |
Assumption register
For each assumption, record: the value, the reason, the source, the owner who can confirm it and the effect on the result if it is wrong. Mark each as confirmed, estimated or unknown.
Validation techniques
- Nameplate check. Compare model data with photographs of equipment.
- Typical value check. Compare with typical ranges for similar equipment. Flag outliers.
- Measurement check. Compare load flow or voltage drop results with measured values at selected points.
- Reasonableness check. Hand calculation at a few points to confirm the software.
- Peer review. A second engineer reviews inputs, not just results.
Sensitivity
Vary the least certain inputs through a plausible range, such as utility fault level, motor starting current or cable length, and see if the conclusion changes. If it does, the input needs better data before the decision.
Reporting
- Summarise inputs and assumptions in an appendix.
- State the operating cases studied and those not studied.
- Give a clear conclusion with the margin and the key sensitivity.
- Keep model files and version them.
Illustrative example
A short-circuit study assumes a utility fault level from an old letter. A quick check with the utility finds the level is now higher after a new substation. The study is rerun, and one feeder breaker is found to be under-rated. A sensitivity case for future growth in fault level is also added to the report. See utility fault level.
Garbage in, confident garbage out
Software produces precise-looking numbers from imprecise inputs. A study is only as good as its data and the honesty of its assumptions.
| Input | Preferred source | Fallback and risk |
|---|---|---|
| Utility fault level | Written statement from the distribution company | Estimate; state conservative case |
| Transformer impedance | Test certificate | Nameplate or typical value |
| Generator reactances | Manufacturer data sheet | Typical values; flag in report |
| Cable lengths | Measured on site | Drawing lengths, often short |
| Load data | Logged measurements | Nameplate with demand factors |
| Protection settings | Downloaded from the relay | Paper records, possibly outdated |
Assumptions register
Keep a list of every assumption with an owner, a source and a date to verify. Each assumption should also state the effect if it is wrong, and whether the conservative direction was chosen.
Validation techniques
- Compare computed load flow with measured voltages and currents.
- Compare computed fault level with utility statements or past records.
- Run a sanity check by hand on one bus.
- Peer review by a second engineer.
- Sensitivity runs on the least certain inputs.
Version control
Identify model version, date, data revision and software version on the report. See brownfield study.
Common mistakes
- Unrecorded assumptions.
- No check of nameplate data.
- One-case studies with no sensitivity.
What to send us
Send the data you have and flag what you do not. Send your application details.
Questions engineers ask
What if data is missing?
Use typical values, label them as assumed and test sensitivity.
How accurate must inputs be?
Accurate enough that conclusions do not change within the plausible range.
Who owns assumptions?
The person who can confirm them, usually the client or equipment owner.
Should I keep model files?
Yes, with version control, so studies can be repeated after changes.
This guide gives general principles. Every study must be reviewed by a qualified engineer.