Solar farm design software: how to choose without replacing what works
The realistic question is not which tool to standardise on. It is which layer of your stack is costing you the most, and whether one tool can take that layer over without breaking the others.
Utility-scale teams do not run one application. They run a CAD-anchored layout tool, a yield engine whose report third parties accept, a spreadsheet for the economics, and often a separate feasibility tool early in development. You can see it in what employers ask for: across 27 job postings for design and project engineers at developers, IPPs and owner’s engineers that we read in September 2026, PVsyst appeared in 15, AutoCAD in 15 and PVcase in 9, usually as a requirement and frequently together in the same posting. Any vendor telling you to replace all of that at once is selling, not advising.
This guide is written by Heliosolve. It is not an independent ranking, and we say where we do not fit.
The four layers, and what “better” means in each
| Layer | The deliverable | What makes one tool better here |
|---|---|---|
| Site and layout | A buildable arrangement on real ground: pitch, GCR, roads, terrain, civil constraints | Whether the geometry survives contact with slope, setbacks and grading — and how fast you can iterate it |
| Electrical design | Strings, inverters, conductors, protection, single-line diagram | Whether the assignment is optimised or merely automatic, and whether the result passes code screening |
| Energy yield | kWh/yr, losses, P50/P90 | Whose report your buyer, IE or lender will accept — and how well the physics holds on bifacial, trackers and terrain |
| Economics | LCOE, NPV, DSCR, debt capacity | Whether it reads the design you actually simulated, or a simplified copy of it |
Most disappointment with design software comes from buying strength in one layer and assuming it transfers to the next. It does not.
What the published evidence says before you compare vendors
Three measurements are worth knowing before any demo, because they set the expectations a vendor cannot.
Tools disagree by a few per cent even on the easiest case. In a blind study, seven commercial tools (3E SynaptiQ, PlantPredict, PVsyst, RatedPower, SAM, SolarFarmer and Solargis Evaluate) modelled two fixed-tilt, monofacial systems with the weather data supplied. Mean error in annual yield ran up to 2.5 % on a 15.4 kW laboratory system and 6.0 % on a 14.5 MW plant (Deville et al., Solar Energy 304, 114207, 2026). Heliosolve was not among the participants, the study does not rank the tools by name in any form we can quote, and no bifacial modules, trackers or terrain were tested. For those there is no published baseline at all.
The person matters more than the tool. Given identical inputs for the same plant, five to seven professional assessors produced P50 values from 1,095 to 1,405 kWh/kWp, and the transposition step alone varied +18.8 % to +24.6 % “even if the yield assessors all claimed to have used the Perez model” (IEA-PVPS T13-18:2020).
Data beats models. Validating an open-source pipeline against well-characterised systems, the authors concluded that “system-specific data availability is more important than model selection” (Deville et al., Prog. Photovolt. 32(5), 2024).
In practice, distrust any accuracy claim that is not tied to a named dataset, a period and named conditions. Ours included. And when you compare two of your own reports, start with why estimates differ.
Evaluate the whole workflow
| Area | The question to make them demonstrate | Evidence to ask for |
|---|---|---|
| Fit | Does it handle our project type, size and stage? | A worked example at our scale, with stated limits |
| Automation | What does the user configure and what does the software generate? | An unedited walkthrough showing every user intervention |
| Review | Can we understand, correct and defend the case? | Identifiable inputs, assumptions and traceable outputs |
| Coherence | Do the energy and economics use the same design? | The configuration that produced both, side by side |
| Handover | What must we still produce elsewhere? | Real deliverables: drawings, schedules, exports |
| Iteration | What happens when one parameter changes? | Do it live — this is where tools diverge most |
| Commercial | What does the offer include, and for how many users? | Current written terms |
Separate contractual requirements from preferences. A deliverable a buyer demands outranks a feature you will admire once.
Test it on one task, with your own project
A prepared demonstration tells you little about your own work. Pick a task you did last month and redo part of it, on one site, with one decision, in one afternoon. Change a condition and explain the result out loud. Write down where you needed help and which numbers you could not trace back to an input.
And be careful with time claims, ours included: an edited demonstration does not establish time savings. If time is the argument, measure it on comparable tasks, and count the preparation, the waiting and the corrections as well as the run itself.
Ask for evidence that matches your work
A validation is only as broad as its dataset. Ask which system, which period, which conditions and against what reference. An annual comparison at one site says nothing about a tracker plant on a slope in a different climate. Sandia’s validation procedure is built around comparing against field data across conditions rather than on annual aggregates (PVPMC). That is a useful frame for the questions, and not a certification of anyone in this guide.
Where the models are demonstrably weakest today is public knowledge, so use it as a checklist: the rear side of bifacial arrays (relative RMSE around 21–23 % versus about 7 % on the front face), module temperature on glass-glass modules, sub-hourly clipping, and tracker backtracking over undulating terrain. All four are documented in why estimates differ.
Cost the whole workflow, licence included
Ask for an offer that matches your users, your projects and your deliverables, then add what the licence does not cover: training, data preparation, review time, and the tools you must keep anyway. Two industry notes:
- Published prices are rare in this category. Most utility-scale vendors quote on request, which makes comparison harder and is itself a data point about the relationship you are entering. Our pricing is published.
- Do not compare a plan for three users against one for fifteen because the sticker is similar.
The three pairwise comparisons
Each of these pages states what the other tool documents publicly, on a stated date, and where we do not have an equivalent:
- Heliosolve and PVsyst — the yield-report layer, and why “replace PVsyst” is the wrong framing for most teams.
- Heliosolve and PVcase — CAD-anchored layout versus a browser workflow that continues into economics.
- Heliosolve and RatedPower — the closest comparison: browser, terrain to finance, and where the differences actually are.
When it makes sense to evaluate Heliosolve
We fit a specific case: a developer engineering team that wants to search layouts against levelized cost of energy, automate string-to-inverter assignment, and follow one design through to energy and financial results in the same place, then review all of it. If your bottleneck is producing the report a lender already accepts, we are not the answer to that layer today, and the honest move is to keep what works and test us on the layer that hurts.
Further reading
A decision you can defend sounds like this: “We kept PVsyst for the yield report, moved layout iteration and the economic comparison to one tool, and here is the task we tested before deciding.”
Heliosolve limits and next step
Energy yield and financial returns are estimates under your project assumptions. Our documented validation against measured data covers one plant; it does not validate every model or location. We have no track record of lender acceptance, and we are a small team. The software does not replace engineering review.
Talk to an engineer. Bring one real task from last month rather than a shortlist of brands.
PVsyst, RatedPower, PVcase, SolarFarmer, Solargis, PlantPredict, HelioScope and SAM are trademarks of their respective owners. Heliosolve is not affiliated with, endorsed by or sponsored by any of them. Product names are used only to identify the products compared.