A model is an engineering claim about how a power system behaves under defined conditions. The claim is only as reliable as its topology, equipment parameters, load and generation data, control settings, study method and validation evidence for that feeder and source date. A polished one-line diagram can still produce misleading results when those inputs have unknown dates or silent defaults.
The buyer's task is to procure a controlled model and a decision-ready study, not simply a software file. The scope should define which questions the model must answer, which operating cases matter, how uncertain data will be handled, what validation is possible and which files allow another qualified engineer to reproduce the result.
What power system modeling services should deliver
Power system modeling services should deliver 4 connected products: a source-data register; a versioned distribution model; documented study cases; and an engineering report that ties findings to inputs. Assumptions need their own register. The package must expose validation limits and software settings, so simulated results are not presented as measured certainty.
Start with the distribution decision. One utility may be checking feeder voltage and thermal loading before an extension. Another may need fault duty and protection coordination for a make-ready change. Distributed generation can add reverse-flow or control questions, while a planning study compares when reinforcement becomes necessary. A shared base model can support these decisions only when its detail and operating cases fit each one.
The Electric Power Research Institute describes OpenDSS as the open-source implementation of a distribution-system simulator for electric utility distribution systems. EPRI lists snapshot power flow plus daily and yearly power-flow modes. Its other named modes include harmonics and dynamics as well as fault study and Monte Carlo fault study. A mode is not a finished study. The engineer still has to select data, cases and acceptance criteria.
Model scope should state the feeder boundary and the equivalent source beyond it. Switch position can change the answer. So can voltage-control equipment within the modeled boundary. Lines and distribution transformers need the phase detail required by the study, while loads or generators need a declared aggregation basis. When a device is intentionally simplified, we record the reason and the approved use of that simplification.
Create a source-to-element lineage
For every model element, record where the value came from and when it applied. The register also identifies who transformed it, the units and the review status. GIS or asset records often establish topology, while interval measurements inform operating cases. Drawings and manufacturer data may fill equipment details. When two records conflict, preserve both entries until the resolution has an owner; convergence is not permission to choose the easier value.
Our utility engineering services workflow separates supplied, measured, calculated, inferred and assumed values. That classification matters during review. A conductor code supplied by GIS and a conductor code inferred from a scanned drawing may share the same text, yet they deserve different confidence and validation actions. The model should retain that difference until evidence closes it.
Set the minimum data by study question
Required data changes with the question. A power-flow case begins with connected phase topology and credible impedance, then adds the loads and controls needed to reproduce the operating condition. Fault analysis depends on sequence and grounding representation that a planning screen may not carry. Protection coordination reaches farther into device settings and sensing. For harmonics, EPRI's OpenDSS documentation begins with a converged snapshot power flow, then applies spectra to harmonic-producing elements and solves selected frequencies. Load assumptions affect resonance sharpness. We therefore review source spectra and the frequency behavior of relevant equipment instead of treating a mode change as a complete harmonic model.
Time resolution also follows the decision. A peak snapshot can screen conductor loading. It may still miss a regulator that operates repeatedly during solar variation. Before committing to an annual simulation, we ask what duration or recurrence would change the recommendation and whether the available shapes can support that answer. Sensitivity cases are then aimed at the uncertainty that matters rather than added for volume.
The Rural Utilities Service states in 7 CFR 1724.40 that borrowers shall have ongoing, integrated planning for additions, improvements, replacements and retirements. The section identifies long-range engineering plans for 10 years or more and construction work plans for 2 to 4 years. It also directs readers to part 1710, subpart F and names RUS Bulletins 1724D-101A and 1724D-101B as additional guidance. The lesson is narrower outside that scope. Name the model horizon and the decision it supports.
| Study view | Inputs that usually control | Decision output |
|---|---|---|
| Power flow and voltage | Topology, phase, impedance, load, generation and controls | Voltage profile, loading, losses and control response |
| Short circuit | Source strength, sequence data, grounding, connections and fault type | Fault current and equipment-duty comparison |
| Protection coordination | Fault results, device curves, settings, sensing and operating sequence | Coordination margins and setting recommendations |
| Time series | Load and generation shapes, control logic, switching and calendar cases | Duration, recurrence and operating-pattern findings |
| Planning alternatives | Forecasts, project options, contingencies and cost inputs supplied by owner | Comparative need, timing and technical effect |
The table is a scoping bridge, not a menu that every project must order. A buyer should ask which requested output the available evidence can support. Sometimes a simple screen is enough to eliminate an alternative; it should not be presented as the final study for a design decision. We identify gaps during proposal development, especially when feeder records require extraction or field verification before a responsible conclusion date can be set.
Define case names that carry meaning
Case names should survive handoff. Use case names that identify horizon, season or date, load condition, generation condition, topology, contingency and project option. “Case 2 final” is not durable. A case manifest should list the parent model version, data snapshot, overrides, disabled controls, solved state and purpose. This prevents an attractive plot from being copied into a report after its assumptions have been replaced.
Scenario counts can grow quickly, so use a decision tree. Begin with base cases, screen credible extremes, identify violations, then test corrective options and sensitivities that could reverse the recommendation. Our distribution line extension design process uses alternatives to compare conductor, phasing, voltage control, route and source options rather than running an undirected collection of simulations.
Validate topology and behavior before reading the plots
Validation begins before the first plot. Connectivity and units are checked as the model is assembled, which catches open islands or devices mapped to the wrong phase. We then reconcile feeder totals and equipment against the source snapshot. Where suitable measurements exist, the comparison uses the same topology and operating time as the model case. Finally, an engineer familiar with the distribution system reviews abnormal results and control behavior for physical sense.
Convergence is only a solver result. A model can still describe the wrong feeder. Consider a regulator placed on the wrong side of an open switch: the voltage profile may look smooth even though the modeled operating path is impossible. Balanced load assumptions can hide a phase problem, and an unsuitable conductor library can shift impedance without producing a solver error. We label the model’s intended use so a planning representation is not quietly reused for service-voltage design.
- Build integrity: test connectivity and normal switch state while reviewing phase continuity, units and equipment-library mappings.
- Operating comparison: reconcile feeder totals first, then compare voltage or device behavior with measurements that represent the same case.
- Clean rerun: execute the released files in a separate environment with the documented software version and preserve the resulting log.
Measurement comparison needs context. Interval data and temporary monitors do not necessarily share clocks or sampling periods, and missing records can bias a seemingly close match. Before calibration, we confirm that the measurement came from the topology and date represented by the case. Changing one feeder load multiplier until the total agrees can conceal phase or control errors, so the comparison remains disaggregated enough to test the behavior under review.
Use sensitivity to expose weak inputs
Sensitivity analysis is most useful when an uncertain input can change the decision. We choose a defensible range for that input, such as load forecast or source impedance, and observe whether the preferred distribution action changes. Settings can matter just as much. Switching assumptions receive the same treatment. If the recommendation reverses, the report identifies the field evidence or operational approval needed before release rather than burying the instability in an appendix.
NERC MOD-032-1, Data for Power System Modeling and Analysis, sits outside this distribution and make-ready service scope. We retain it only as a contrasting governance reference because it formalizes data requirements, case definitions and procedures for resolving technical issues within a controlled study process. It does not govern an ordinary distribution study automatically, and we do not use it to expand a feeder assignment beyond the owner’s stated boundary.
Report findings as conditional engineering decisions
Results need operating context. Each finding should state the case, criterion, result, location, severity, data confidence and recommended action. “Low voltage exists” is incomplete. A useful record identifies which operating case produced the value, which criterion was used, whether controls operated, how persistent the condition is and which input uncertainties could change it. The report should distinguish observed data from simulation output throughout.
Protection work deserves careful boundaries. A short-circuit model supplies fault-current inputs, but device coordination also depends on interrupting ratings, sensing, settings, curves, operating philosophy, distributed generation, arc-flash study assumptions where relevant and utility practices. Our protection coordination study engineering scope defines those interfaces rather than promising that one fault run resolves every device decision.
Recommendations need a field consequence. They should be testable against the model and the next design step. A reconductoring proposal identifies the affected segment and the case that drives the need, then records the assumed conductor and expected voltage or loading change. For a regulator setting, the report shows the control mode and coordination effect before requesting operational approval. Generic upgrade lists are omitted because they cannot be traced to a case result.
Self-critical note: A detailed model can create precision beyond its evidence. We flag inferred equipment, stale load data, unverified switch states and uncalibrated control behavior in the same view as results. When an input can reverse the decision, the report calls for verification instead of adding decimal places.
Deliver files that another engineer can rerun
A reproducible package begins with the released source register and equipment library. The base model is accompanied by its case overlays and run instructions, including the exact software version. Logs and result exports show what the run produced; plots and the report explain what the engineer concluded. We include transformation scripts when data was processed and keep the exception history with the package. Credentials and restricted personal data are removed before handoff.
Model ownership and update duties should be explicit. Decide who may edit the base, who approves library changes, how field and as-built updates enter, how study branches merge and when the model is revalidated. A project model that is never maintained can still be a valid dated study artifact. It should not be presented as a live operating model after its source snapshot expires.
Choose power system modeling services with release criteria
Utility planning teams: frame the assignment around the decision and the operating cases that could change it. When source gaps remain, we release the work in stages rather than treating an early model build as an accepted recommendation. Each report finding names the uncertainty that still limits action.
Asset and protection teams: require element-level lineage and a clean rerun package. Before fault or coordination findings move forward, our engineers reconcile the device record and resolve any switch state or setting that could alter the result. This review provides a specific hold point instead of a general warning about data quality.
The request should name the distribution question, the feeder boundary and the study horizon. It also identifies the cases and criteria the owner expects, together with the software or file constraints that affect handoff. Supplied data can be inventoried at kickoff, but expected gaps need their own response path. Assumption approval and control-setting authority remain with named utility roles before final runs begin.
Our in-house engineering team performs distribution and make-ready modeling, engineering and design. When a modeled recommendation advances into full turnkey distribution or make-ready construction, Draftech-managed subcontract crews deliver that work under Draftech's QA/QC and safety program at the owner-approved distribution sites. The vendor coordination process carries field feedback into controlled design review without blurring engineering responsibility or expanding the assignment into a different power scope.
Acceptance starts with a model that can be traced back to its source snapshot. Critical data gaps must either be resolved or bounded in the report. Automated checks and the clean rerun then establish that the released files behave as documented. Peer comments stay open until a named reviewer closes them. If the owner accepts a remaining exception, the gate records its effect and that decision instead of calling the model unconditionally complete.
Commercial terms should follow the same sequence. Data intake establishes what was received, not whether every record is usable. The next milestone can cover model assembly and its exception register, while study acceptance waits for the review log and released run files. Owner decisions about assumptions or operational settings become explicit hold points, protecting the schedule from an analysis that cannot responsibly proceed.
Security and access belong in the scope as well. Distribution models can reveal facility locations and operating details, so the owner defines the approved storage and transfer path before data arrives. Access is limited by role, with retention and deletion handled under the same plan. We review those controls separately from technical validation; a correct result is not ready for delivery through an unauthorized channel.
A useful handoff leaves the next engineer with a clear starting point. That engineer can rerun the released case, see which feeder condition drove the recommendation and identify the input that still needs verification without reconstructing private assumptions. When load or topology changes later, the dated model remains an auditable study rather than being mistaken for a live system record. The buyer can then update the right layer instead of rebuilding the reasoning from a plot.
Send the current one-line and the decision that cannot proceed, or email Draftech for secure intake. Our first response will identify the source records that need validation and the owner decisions that must remain outside the model.
Modeling scope ends at the study. The distribution or make-ready route that follows can begin at no cost: Draftech's free permitted design engineers the first 20,000 linear feet of a qualifying route through permit approval, with owner assumptions and settings authority left exactly where this article puts them.
