Six practical points when performing sensitivity analysis in PVSyst

Table of contents

• Why sensitivity analysis becomes important in PVSyst

• Practical point 1: Fix the purpose and comparison axis of the sensitivity analysis up front

• Practical point 2: Don’t change too many conditions at once

• Practical point 3: Define a baseline case clearly and read the differences

• Practical point 4: Set realistic ranges for the conditions you vary

• Practical point 5: Look not only at annual values but also at monthly breakdowns and loss structure

• Practical point 6: Link results to design changes and on-site checks

• How to connect PVSyst sensitivity analysis to practical decisions

Why sensitivity analysis becomes important in PVSyst

When you run photovoltaic simulations with PVSyst, the figures that initially draw attention are often annual energy production and PR. Of course those are important indicators, but in practice you rarely make final decisions based on a single assumed condition. During design and planning there are multiple unresolved conditions—azimuth, tilt, shading conditions, how the PCS is configured, loss coefficients, utilization losses, and so on. Sensitivity analysis is a practical exercise to check how much the results move when such uncertain conditions change and to organize which conditions should be weighted more heavily.

In practice, sensitivity analysis is not useful simply for finding the highest numerical result. Its real value is in understanding which conditions have a strong impact on outcomes and which conditions can move somewhat without materially changing overall judgment. For example, if a small change in row spacing greatly alters annual energy, that condition is very important for design. Conversely, if adjusting a certain loss coefficient barely affects the results, that condition can be given low priority. PVSyst sensitivity analysis is a tool for identifying design priorities.

Sensitivity analysis is also effective for strengthening internal explanations and proposal materials. Saying “this option is the best” alone may not convey sufficient rationale to the reader. But explaining that one option is sensitive to shading while another remains stable when PCS conditions change, or that some options show little variability depending on how loss coefficients are set, elevates the quality of the comparison. In practice, more than the absolute numbers, stakeholders often care about how stable those numbers are and how much they change when assumptions move.

Furthermore, careful sensitivity analysis clarifies what to check next. Conditions that cause large result swings should be prioritized for site checks or additional design work, while conditions that hardly move the results can remain provisional early on. In other words, PVSyst sensitivity analysis is not about producing more calculations but about organizing decision-making efficiently. Below, the following six points outline how to make sensitivity analysis useful in practice.

Practical point 1: Fix the purpose and comparison axis of the sensitivity analysis up front

The first thing to do when performing sensitivity analysis is to clarify why you are analyzing. In practice there are many conditions you care about, and it’s tempting to check everything—azimuth, shading, PCS capacity, loss coefficients, etc. However, if you start analyzing in that state, you’ll generate many numbers but the conclusions tend to blur. Sensitivity analysis is not an exercise in freely changing conditions; it’s important to decide beforehand what you are comparing to verify.

For example, whether you want to see how much difference row spacing makes to annual energy, how much azimuth deviation affects results, or how PCS condition changes affect output limitation will determine which results you should look at. Whether you use annual energy as the main metric, PR, or seasonal monthly differences will also depend on purpose. If this isn’t fixed, the more results PVSyst produces, the more confused you may become, and you might not know which option is stable.

Also, narrowing the comparison axis to one factor is very important. In one comparison you might look only at shading differences, in another only at DC/AC ratio—partitioning the issues makes it easier to interpret differences. In practice, changing multiple assumptions at once can feel more realistic, but it mixes the reasons for differences. As a result, you may not know whether the effect came from shading, PCS settings, or climatic assumptions.

Therefore, before starting sensitivity analysis, write a one-sentence decision on what to vary and what to fix in that comparison. For example, “Check the impact of different tilt angles on annual energy” or “Look at winter loss differences when row spacing is changed.” Doing this helps you read PVSyst results without ambiguity. The first step to a successful sensitivity analysis is not increasing the number of analyses but fixing the purpose and axis.

Practical point 2: Don’t change too many conditions at once

A common mistake in sensitivity analysis is changing too many conditions at once. In practice, wanting a more realistic comparison can lead you to change azimuth, tilt, PCS capacity, loss coefficients, shading, and so on simultaneously. The numbers that result may be meaningful as a single scenario, but it becomes hard to tell which condition contributed how much. The true purpose of sensitivity analysis is to find the influential conditions, so changing too many variables at once blurs that purpose.

For example, if you slightly alter azimuth, change tilt, widen row spacing, and adjust PCS settings and then compare that scenario to a baseline, you will see differences. But it becomes difficult to determine whether those differences came from azimuth, shading improvements, or PCS settings. What you really want in practice is to know which part of the design most strongly affects the results. For that, you need comparisons that show how results change when a single variable is varied.

Changing multiple conditions at once also reduces reproducibility. Later, another engineer looking at the results may have trouble tracing why differences occurred, and even you may find you need to reorganize assumptions when re-running analyses. PVSyst handles many conditions, so if you don’t manage them carefully, you end up increasing workload rather than analysis quality. Sensitivity analysis is not a numbers race; it’s about clarifying the meaning of differences.

Of course, at the final stage you also need scenario comparisons where multiple conditions move simultaneously. But do that after you’ve checked single-variable sensitivities to some degree. First verify the effect of one variable, then examine combinations—this sequence stabilizes your reading of differences. If you want PVSyst sensitivity analysis to be practical, change one thing at a time and build meaning cumulatively rather than changing many things at once.

Practical point 3: Define a baseline case clearly and read the differences

To make sensitivity analysis useful in practice, defining a baseline case is indispensable. The baseline case is the starting scenario for comparisons. When conducting sensitivity analysis in PVSyst, if the baseline is ambiguous, interpretation of differences will be ambiguous too. If you use Plan A as the baseline in one comparison and Plan B in another, you may see increases or decreases in numbers but have difficulty organizing which plan is stronger where.

The baseline case does not have to be the best plan. In practice it is often a natural plan, a plan relatively consistent with site conditions, or a plan that has already been shared internally. The important thing is that the baseline is fixed. If the baseline wobbles, the meaning of the same change can look different. What you want from PVSyst sensitivity analysis is information on “how much the result changes if this condition is changed for this plan,” so if the starting point is not fixed the analysis axis drifts.

Also, having a clear baseline makes it easier to explain differences. For example: “Widening row spacing by 0.5 m (1.6 ft) from the baseline reduced winter shading losses and improved annual energy by X%” or “Changing PCS capacity from the baseline reduced summer output limitations, moving PR by Y%.” This is very effective for internal explanations and proposals. Instead of simply saying an option is higher or lower, you can show how it changed from the baseline.

In practice, it’s convenient to choose one baseline at the start and keep a fixed document of that scenario’s settings. If you know what assumptions the baseline uses, later comparisons are easier to reproduce. When running sensitivity analysis in PVSyst, the clarity of the baseline is more important than the differences themselves for correctly reading the results.

Practical point 4: Set realistic ranges for the conditions you vary

An often underestimated factor in sensitivity analysis is how wide you vary each condition. Theoretically, varying a parameter widely makes changes easier to see. However, what matters in practice is knowing how much results move within ranges that can realistically occur. If you only look at extreme setting differences in PVSyst, the numbers may move significantly but those results are hard to use for practical decisions.

For example, the meaning of production differences when you change azimuth slightly versus when you introduce an angle difference you would never adopt on site is different. The same applies to row spacing and tilt. You should vary parameters within the range that can actually be chosen given site and earthwork constraints. Comparing extreme conditions that won’t occur in practice rarely informs design prioritization. Sensitivity analysis should be used to understand sensitivity to realistic field variations, not to find theoretical maximum differences.

Likewise, varying loss coefficients or utilization losses outside of practical operational sense weakens the comparison. If one scenario treats temperature losses extremely lightly while another is highly conservative, the observed differences may look large but it’s hard to tell whether they reflect design differences or mere assumption differences. When conducting sensitivity analysis in PVSyst, keep ranges tied to site intuition.

Therefore, before changing values, check whether the chosen ranges are likely in practice and whether they fall within site constraints and decision boundaries. Looking at differences within realistic ranges makes the numbers directly relevant to design decisions. In PVSyst sensitivity analysis, larger ranges are not better—realism is.

Practical point 5: Look not only at annual values but also at monthly breakdowns and loss structure

When reading sensitivity analysis results, it’s important not to look only at annual values but also to check monthly generation and the loss structure. After running sensitivity analyses, it’s tempting to judge by annual energy or PR differences first. Of course annual values matter for final comparisons. However, if you don’t check which months the differences concentrate in and which loss items create those differences, you won’t know what to improve.

For example, even if the annual difference is small, a large winter difference may indicate that shading or tilt has a strong effect. If differences are large only in summer, temperature losses or PCS settings may be dominant. Viewing the loss tree reveals whether the difference arises from shading losses, temperature losses, wiring, or utilization losses. PVSyst sensitivity analysis is useful not only to compare annual totals but also to know which loss items are sensitive to which conditions.

Also, checking monthly trends and loss structure makes it easier to explain the character of each option. One option may be only slightly different on an annual basis but more stable in winter and stronger against shading. Another option may be stronger in summer and have more headroom on the PCS side. Since projects prioritize different seasons and manage risks differently, judging with monthly tendencies as well as annual totals creates stronger comparisons.

As a practical countermeasure, for each sensitivity case always return to the monthly graph and the loss tree after checking annual energy. Confirm which months show differences and which losses moved—the meaning of results becomes much clearer even with just those checks. To make PVSyst sensitivity analysis practical, don’t be satisfied with annual values alone; follow through to where and how differences appear.

Practical point 6: Link results to design changes and on-site checks

The final point is to connect sensitivity analysis results to design changes and on-site verification. Sensitivity analysis is not an exercise that ends with looking at numbers. Its real purpose is to identify which conditions the results are sensitive to and what should be prioritized for refinement. If changing a condition in PVSyst produces a large difference, that condition should be prioritized for on-site checks or additional design work.

For example, if a small change in row spacing greatly moves winter generation, prioritize verifying shading conditions for that project. If a slight change in PCS settings significantly alters summer PR, it’s worth refining assumptions like DC/AC ratio and output limitation. If slight adjustments to loss coefficients move annual values a lot, re-examine the validity of those loss assumptions. Sensitivity analysis is a tool to decide what to verify next.

Linking analysis to on-site checks also helps confirm whether desk-calculated differences are meaningful in the field. For shading-sensitive options, focus site checks on building positions, slope directions, and morning/evening obstructions. If ventilation conditions matter, re-evaluate spacing density and surrounding structures. Returning PVSyst sensitivity analysis results to the site turns a mere comparison into work that improves design accuracy.

In practice, don’t stop at summarizing sensitivity results in tables and graphs; organize them into a set of “conditions that showed large differences,” “conditions needing additional verification,” and “conditions that should be prioritized for on-site checks.” This directly connects PVSyst results to next actions. The key to making sensitivity analysis practical is converting numeric comparisons into prioritized design changes and site checks.

How to connect PVSyst sensitivity analysis to practical decisions

What the six points above share is the principle of not making sensitivity analysis mere number play. Fix the purpose and comparison axis, avoid changing many conditions at once, clarify the baseline case, vary conditions within realistic ranges, check monthly breakdowns and loss structure as well as annual values, and finally connect results to design changes and on-site verification. Following this flow makes PVSyst sensitivity analysis a practical analysis that supports design decisions rather than just a calculation exercise.

What really matters to practitioners is not finding the option that produces the highest number. It’s determining which conditions are most sensitive, where uncertainty is largest, and which aspects can vary without changing judgment. Sensitivity analysis helps not only reduce numerical uncertainty but also clarify decision priorities.

Improving the quality of sensitivity analysis also means not stopping at desk comparisons. If site conditions are ambiguous, no amount of careful assumption variation in PVSyst will make the differences meaningful. Only when you tie results to the actual site—walkways, slopes, obstructions, and maintenance access—does the sensitivity indicate real operational risk or room for improvement. Use PVSyst as a tool to connect site conditions to calculations.

In that sense, when you want to improve the certainty of on-site position checks or coordinate acquisition, using an iPhone-mounted high-precision GNSS positioning device such as LRTK can be effective. If you can organize the positional and site condition information obtained on site, the assumptions for PVSyst sensitivity analysis become clearer and it becomes easier to judge which condition differences truly matter. When PVSyst improves desk-level comparison accuracy and LRTK supports on-site accuracy, sensitivity analysis becomes not just verification of numerical changes but a path to on-site–rooted practical decisions. Conducting careful PVSyst sensitivity analysis thus not only increases simulation precision but also cultivates the design capability that bridges desk work and the field.