Five Causes Why PVSyst Generation Forecasts Diverge and How to Review Them

Many practitioners struggle when PVSyst generation forecasts do not match actual generation. Even input conditions that seemed reasonable at the design stage can show large discrepancies when compared to operational results, making it hard to know where to start investigating.

However, forecast divergence is not simply caused by low software accuracy. In most cases there is room for review in how meteorological data are selected, how design conditions are entered, how shading is handled, how losses are set, and how comparisons with actual results are made. In other words, using PVSyst results correctly requires not only looking at the calculated outputs, but carefully retracing the assumptions under which the calculations were made.

In practice, generation forecasts directly affect multiple decisions—internal explanations, client briefings, construction planning, financial checks, and post-commissioning root-cause analysis. If forecast deviations are left unaddressed, they can dull not only judgment on plan validity but also on site improvement decisions. That is why it is necessary to grasp the structure of the deviations and organize how to review them.

This article categorizes five representative causes of generation forecast divergence in PVSyst, and clearly explains the checkpoints for each and how to review them in practice. It is useful both for those about to create forecasts and for those already facing gaps between forecast and operational results.

Contents

• Key concepts to grasp first when PVSyst generation forecasts diverge

• Cause 1: Meteorological data assumptions do not match actual conditions

• Cause 2: Missing or weak input values for design conditions

• Cause 3: Shading treatment does not match site conditions

• Cause 4: Loss settings are more optimistic than real operation

• Cause 5: The method of comparing post-operation results is off

• Procedures to improve PVSyst forecast accuracy

• On-site measures to reduce forecast divergence

• Summary

Key concepts to grasp first when PVSyst generation forecasts diverge

When PVSyst generation forecasts diverge, the first thing to understand is that forecast values are not numbers that simply predict the future, but results calculated under a set of assumptions. In other words, the reported annual or monthly generation values are valid on the premise that the entered meteorological conditions, equipment conditions, loss assumptions, and operating conditions are appropriate. If the assumptions are even slightly off, the results will naturally deviate.

It is important not to jump immediately to a single cause the moment you see a difference between forecast and actual values. In practice, attention tends to focus on single points—perhaps there was less irradiance, perhaps equipment underperformed, perhaps there was an input error—but in reality, several small differences often accumulate to create an overall divergence. If the meteorological data are a bit dated, shading conditions are simplified, loss rates are set more lightly than reality, and the comparison period includes downtime, these factors combined can produce a non-negligible annual difference.

Also essential is aligning what is being compared. If the PVSyst output being viewed is on the AC side or DC side, whether the actuals are based on revenue-metering or internal monitoring, and whether the comparison period is the full year or a partial period—if these are misaligned, they can lead to incorrect conclusions. Reviewing generation forecasts involves not only recalculating numbers but also clarifying what is being compared to what.

Furthermore, forecast deviations are not entirely negative. They are an opportunity to reveal conditions missed at the design stage or site characteristics that only become apparent after operation begins. By checking which months show the largest gaps, whether the bias is morning or afternoon, or whether trends differ between hot and cold seasons, you can narrow down likely causes. Treating forecast-versus-actual differences as material to improve design and operational accuracy, rather than merely as failures, is a fundamental attitude for using PVSyst in practice.

Cause 1: Meteorological data assumptions do not match actual conditions

One of the most typical causes of PVSyst forecast divergence is a mismatch between the meteorological data assumptions and the site’s actual conditions. Forecasts are highly influenced by solar irradiance, temperature, and wind effects, so discrepancies here greatly affect overall results. Even data that appears to represent a location close to the project site can diverge from actual conditions if terrain, elevation, coastal versus inland location, or seasonal cloud patterns differ.

A common mistake is using representative-year meteorological data to create an annual forecast, then directly comparing those numbers to a single year’s actual results. Representative-year data are based on long-term averages, so if a particular year was sunnier or cloudier than average, it is natural that that year’s actuals will differ. Judging simulation settings as incorrect based solely on a single year’s difference can lead to misdirected revisions.

Local variations can be larger than expected even within the same region. Mountainous areas have different morning/evening shading and fog tendencies; plains can differ in ventilation and temperature rise depending on surrounding environment. Higher temperatures reduce output, and weak winds make it harder for equipment to cool—so temperature differences as well as irradiance differences are sources of forecast error.

When reviewing, first check which location the meteorological data represent and calmly assess geographic differences from the project site. Then compare monthly forecasts and actuals to see in which seasons the gap widens; this makes it easier to suspect meteorological assumption errors. If the gap is large only in summer, consider high-temperature conditions and cloud tendencies in hot months; if only in winter, consider low irradiance or snow effects—season-specific factors will emerge.

To further improve accuracy, interpret site operation results with the actual meteorological tendencies for the year in question. Rather than mechanically matching representative-year forecasts to actuals, determine whether the year was sunnier or less favorable than average before judging differences; this helps separate design-condition issues from meteorological issues. Meteorological data form the foundation of generation forecasts, and as long as that foundation remains vague, refining other conditions will not stabilize accuracy.

Cause 2: Missing or weak input values for design conditions

The second cause is omissions or weak entries for equipment and layout inputs. PVSyst is powerful, but it does not automatically compensate for less-than-complete site understanding beyond the entered conditions. If any assumption—tilt angle, azimuth, system capacity, row spacing, string configuration, the design values underlying loss assumptions—differs from reality, the output will be pulled in that direction.

Although values may be consistent on drawings at the design stage, minor changes often occur during construction. For example, layout shifts due to ground conditions or obstacles, planned spacing cannot be achieved on site, surface conditions after earthwork change the perceived tilt, or wiring routes become longer—such changes are common. If these are not reflected in the simulation, differences will appear between design forecasts and as-built performance.

Be cautious of proceeding without digging into certain items beyond their initial defaults. It is easy to focus on large values during input, but the details are more likely to create practical differences. Slight azimuth deviations, greater-than-expected temperature rise, differences in cable specifications, and real-world operating constraints—small differences can subtly affect monthly and annual results.

The basic review approach is to separate and compare design drawings and the as-built state. Contrast what was entered as planning assumptions with what changed by the completion stage; this clarifies omissions. Particularly when comparing to actuals, check that array conditions and installation conditions in the model match as-built drawings and on-site confirmations.

Also note that weak inputs tend to cascade. For example, lax row-spacing input affects shading treatment, and underestimating temperature conditions affects loss settings. Reviewing design conditions should not be a one-off fix but rather treated as interconnected items—layout, equipment specification, electrical conditions, and operational conditions are linked. Improving PVSyst forecast accuracy requires not just filling input fields, but striving to reflect site reality.

Cause 3: Shading treatment does not match site conditions

The third cause is that shading treatment is overly simplified relative to site conditions. Forecasts depend not only on irradiance but on how much of that light actually reaches the equipment. In practice, the effects of surrounding obstacles or terrain are often underestimated or insufficiently reflected, resulting in lower-than-expected actual generation.

Shading impact is not constant throughout the year. At times and seasons with low solar altitude, even minor obstacles can have a large effect. Conversely, judging only by noon conditions can lead to overlooking morning/evening or winter losses. Nearby trees, slopes, structures, fences, and existing equipment—even if visually negligible on site—can certainly affect generation.

Micro variations in elevation created during development or installation are also not negligible. Although drawings may look regular, on site the relative heights between rows or undulations in the ground can cause shading to occur sooner than expected. The larger the site, the more likely you are to misjudge overall trends if you rely on a single representative section.

When reviewing shading, first clarify which shadows are modeled and to what extent. Are you only considering near-field obstacles, or are you accounting for terrain shading? Have any structures or vegetation changed after construction? Then, look at the time-of-day profile of generation drops; the pattern of declines by time can indicate shading. If only mornings are low, or if gaps are large in winter mornings/evenings, shading assumptions warrant scrutiny.

The key in shading evaluation is tying desk inputs to how the site appears in reality. Combining photos, layout information, height data, and seasonal observations makes it easier to identify which obstacle affects which time periods—not just speculation. PVSyst calculations faithfully follow the input shading conditions but are not strict about what was missed onsite; thus shading treatment is an item where differences in field verification create the biggest variance.

Cause 4: Loss settings are more optimistic than real operation

The fourth cause is that loss settings are more optimistic than actual operation. Forecasts move from ideal generation potential toward real outputs by stacking expected losses. If this loss modeling is too lenient, calculated values may look cleanly high while operational results do not reach those levels.

Common practical loss contributors are soiling, temperature rise, wiring losses, equipment variability, conversion inefficiencies, downtime, and curtailment. Each may seem small in isolation, but over a year they add up. During planning it is easy to assume near-ideal conditions and fail to fully incorporate maintenance frequency or harsh site environments.

For example, soiling depends on region and operation conditions. If a site experiences long dry periods, dust, or conditions that promote buildup, using a uniform low soiling loss will make forecasts overly optimistic. In high-temperature environments, temperature-induced output decline can be stronger than expected; if ventilation conditions and installation details are not sufficiently considered, this often appears as a summer shortfall.

Insufficient consideration of downtime and curtailment is also common. Equipment inspections, communication failures, minor faults, planned stoppages, and grid-side constraints are elements of real operation that are hard to foresee during design. A forecast that ignores these assumes the equipment always operates ideally, making deviation from actuals likelier.

When reviewing, it is effective to inspect loss items separately rather than as a single lump sum. Identify which losses were set large and which were underestimated, and cross-check with months or time periods showing the largest gaps. If the gap widens in summer, consider temperature and soiling; if the shortfall is consistently gradual year-round, consider equipment variability and conversion losses; if only specific periods show large drops, consider downtime or curtailment. Loss setting is not a final tweak but a crucial step to translate field reality into numbers.

Cause 5: The method of comparing post-operation results is off

The fifth cause is that the method of comparing operational results and PVSyst forecasts itself is misaligned. This oversight is surprisingly common: even if simulation conditions are reasonably valid, different comparison methods can make the gap appear large. In other words, sometimes the divergence is due to the comparison method rather than the forecast.

A typical issue is mismatched boundaries for the measured energy. If the PVSyst output and on-site measurement points do not correspond, whether wiring or conversion losses are included will differ. Ignoring differences in metering points and just lining up numbers can produce unexpected gaps. Monitoring data gaps and correction processes can also make actual values appear lower.

How the period is selected matters too. Commissioning months often involve ramp-up adjustments and stoppages, so evaluating only the first month can be overly harsh. Conversely, picking only unusually good-weather months to claim the forecast failed misrepresents the overall picture. Annual, monthly, and daily views reveal different causes.

Also, how abnormal days are handled in the comparison should be clarified. If you include days with clear equipment stoppage, long communication outages, or external constraints in the comparison, evaluation of forecast validity and evaluation of operational troubles get mixed together. Ideally, assessing forecast validity should be separated from extracting operational anomalies.

When reviewing, first align the comparison boundary, period, and data quality. Then, examine not only annual totals but monthly trends, daily tendencies, and, if necessary, time-of-day patterns to identify which issues are central. If comparison methods are not well established, you cannot decide whether to suspect meteorology, design, shading, or losses. In many cases, reviewing PVSyst should start with organizing comparison rules before re-entering simulation inputs.

Procedures to improve PVSyst forecast accuracy

To improve PVSyst forecast accuracy, do not randomly adjust numbers; instead, follow an ordered review sequence. The first step is to align comparison conditions between forecast and actual values. Without clarifying which period, which metering point, and what data quality are being compared, you cannot determine whether a change to inputs improved or worsened results.

Next, check monthly differences. Looking only at annual totals blurs the cause, but arranging differences by month reveals seasonality. Whether performance is weak in summer, weak in winter, consistently low year-round, or poor only in specific months changes which items you should suspect. Summer shortfalls point to temperature or soiling; winter shortfalls point to shading or low-irradiance conditions; consistent year-round shortfall increases the likelihood of loss settings or metering boundary issues.

After that, review meteorological data. Confirm whether the assumptions used are appropriate for the project site and whether there is a reasonable interpretation when comparing to single-year actuals. It is important not to conflate meteorological effects and design effects; treating a meteorologically unfavorable year as a design problem can lead to unnecessary corrections.

Once meteorological factors are clarified, review design inputs and shading. Check that modeled values match as-built conditions and on-site verification results, and update inputs if necessary. At this stage, being able to pick up subtle differences not visible in desk drawings is important—small array or height differences can impact shading and temperature.

Finally, align loss settings with actual operation. Using past operating results and maintenance realities, check for overly optimistic assumptions. Rather than making large changes at once, review items one by one so you can trace which correction reduces which gap. Improving forecast accuracy is not about guessing the correct answer once, but about iteratively refining hypotheses and validating them. PVSyst is a powerful tool for that process, but it must be used within an organized review workflow.

On-site measures to reduce forecast divergence

If you truly want to reduce forecast divergence, do not confine yourself to the simulation screen—improving the quality of on-site information is indispensable. In practice, accuracy gaps often stem less from how skillfully inputs are entered than from how reliable the underlying information is. Project location relationships, positions and heights of surrounding obstacles, post-development topography, as-built array status, and organized operational monitoring data—all require solid site information to avoid judgment drift.

Especially important is preventing information fragmentation between the design, construction, and operation phases. If design assumptions change during construction, there must be a process to feed those differences back into the simulation. If specific time-of-day drops appear after commissioning, site obstacles and array conditions should be easily rechecked. Projects that can perform these feedback loops tend to improve forecast accuracy on the second and third projects.

Forecasting is not only the work of drawing staff. Construction managers, maintenance personnel, and on-site inspectors each hold pieces of information that, when connected, produce realistic assumptions. For example, information such as minor layout changes after construction, remaining shading factors nearby, cleaning frequency likely being lower than assumed, and frequent stoppages during ramp-up are often known by those close to the site. Whether that information is reflected in PVSyst assumptions greatly affects the usability of forecasts.

Moreover, to avoid repeating the same divergence in future projects, do not treat forecast-versus-actual differences as one-offs; document them. Accumulating project-level records of which assumptions were overly optimistic, which assessments were weak, and how differences manifested by month will steadily improve your company’s forecasting accuracy. Improving PVSyst usage in practice is most effective when it includes not only software skills but also a system for on-site measurement and post-project review.

Summary

The causes of PVSyst forecast divergence can be broadly organized into five points: meteorological data assumptions, design input values, shading treatment, loss settings, and the method of comparing actuals. When a forecast seems off, do not suspect only one cause; instead, inspect assumptions and comparison conditions in order. Generation forecasting is not a one-time task but a process of improving accuracy through comparison with actuals.

In practice, site-specific position, height, shading, layout, and as-built difference information matter more than desk input values. To make PVSyst reviews reliable, do not treat site information as vague—confirm and record it. Those who want to improve forecast accuracy should not separate simulation and site verification into distinct tasks, but connect them as a single workflow.

In that respect, when you want to efficiently confirm on-site positions and equipment layout or record surrounding conditions, adopting field-friendly tools such as LRTK has value. LRTK, as an iPhone-mounted GNSS high-precision positioning device, pairs well with situations that require high-precision on-site location data, and is useful for verifying shading and layout conditions and organizing post-construction records. If you want to bring PVSyst forecasts closer to site reality rather than leaving them as desk numbers, reviewing on-site measurement accuracy as part of the process will ultimately reduce forecast errors.