Quick answer
An evaluation of modeled or observed PV energy reductions, using defined reference conditions, energy boundaries and evidence to investigate their causes.
Answer: Solar loss analysis investigates where and why PV energy is reduced relative to a defined reference. A design study estimates loss mechanisms; an operating-system investigation compares measured energy with a model using corresponding weather and time periods. Match the electrical boundary, validate the data and examine residual patterns before attributing a shortfall to equipment, maintenance or design.
What is being compared?
A loss figure needs a reference. Nameplate capacity in kW is a power rating under specified test conditions. Monthly generation in kWh is energy. Dividing one by the other does not establish a percentage energy loss.
Distinguish two jobs:
| Analysis | Starting question | Useful evidence |
|---|---|---|
| Design loss study | Which modeled mechanisms reduce estimated production? | Resource data, geometry, equipment, loss assumptions and simulation outputs |
| Operating-system diagnosis | Why does measured production differ from an appropriate expectation? | Meter data, corresponding weather, operating records, model assumptions and uncertainty |
A simulation loss diagram is an accounting view of a model. It does not prove that the operating plant experienced each displayed loss. Conversely, a measured shortfall does not identify its cause by itself.
For retained-fraction arithmetic, use loss factor calculation. For what a multiplier includes, use panel derate factor. This page focuses on investigating mechanisms and reconciling evidence.
Establish the energy boundary first
Record where each result is measured or modeled: array DC output, inverter AC output, plant delivery meter or exported energy after on-site consumption. Identify whether transformers, auxiliary demand, storage and curtailment are included.
An inverter-output estimate compared with an export meter can show a difference even when the system works as intended. On-site consumption is not automatically a conversion loss. For storage projects, also record charging energy, discharge energy and state-of-charge treatment across the reporting period.
Use a comparison register:
| Field | Record before interpreting a difference |
|---|---|
| Energy boundary | Meter location and equipment upstream/downstream |
| Time period | Identical dates, time zone and interval convention |
| Resource | Historical forecast weather or weather corresponding to the measured period |
| Plant configuration | Module/inverter counts, orientation and changes during the period |
| Included effects | Which predictable losses the expectation already includes |
| Data coverage | Missing intervals, invalid readings and substitutions |
| Restrictions | Export limits, outages, commissioning and maintenance periods |
| Uncertainty | Measurement and model limitations relevant to the decision |
Diagnose a shortfall in five steps
1. Validate the records
Check meter units, cumulative-register resets, timestamps, daylight-saving handling and missing intervals. A communications outage can remove monitoring records without stopping generation. Do not replace every absent interval with zero and call it lost production.
Inspect weather observations too. A dirty or misaligned irradiance sensor can distort the expected-energy calculation. Keep the raw data and document corrections, excluded intervals and any replacement weather source.
2. Build an expectation for the same conditions
A forecast based on typical weather answers a different question from a model driven by the weather during the measured month. Comparing a cloudy month directly with its historical forecast can misclassify resource variation as plant underperformance.
The NREL federal-system assessment compares measured production with modeling informed by coincident irradiance and ambient temperature, aligning the data over corresponding intervals. Apply that principle at the resolution supported by your records.
Document predictable effects already included in the model. If expected output already includes wiring and inverter conversion, the measured-versus-expected residual is not another estimate of those entire losses. Adding them again double counts the reference deductions.
3. Examine when the difference occurs
Compare monthly totals, then inspect shorter intervals where valid records exist. A midday plateau, a persistent deficit on one inverter and a step change after maintenance suggest different investigations.
PVPMC’s model-validation procedure combines error metrics, residual analysis and comparison with a baseline model. A residual is the difference between measured and modeled output. Its relationship to irradiance, temperature and time can help identify questions to test, rather than settling causation from one annual total.
4. Test mechanisms against independent evidence
| Candidate explanation | Evidence to seek | Interpretation limit |
|---|---|---|
| Shading | Obstruction survey, shade model, affected strings and time-of-day pattern | A shaded area percentage is not automatically an energy-loss percentage |
| Soiling | Inspection, suitable soiling measurements or a controlled before/after comparison | Weather and operating changes can confound a cleaning comparison |
| Temperature behavior | Module-temperature data or supported thermal model, equipment coefficients | Ambient temperature alone is not module temperature |
| Inverter limitation | AC power trace, inverter status and configured limits | A plateau can reflect clipping, curtailment or another limit |
| Availability | Alarm/event logs, affected capacity and resource during outages | Duration alone does not quantify lost kWh |
| Electrical fault or mismatch | Comparative string/inverter records and qualified inspection | Monitoring patterns suggest tests, not a safe diagnosis by themselves |
| Model or measurement error | Configuration audit, sensor checks and alternative-model comparison | Similar annual totals can hide offsetting errors |
A qualified electrical professional should investigate suspected electrical faults. Do not infer a safe repair procedure from an energy graph.
5. Record an action and verify its result
For each finding, record the evidence, competing explanations, responsible owner, proposed action and follow-up metric. Separate confirmed causes from hypotheses.
Recheck performance using consistent boundaries and suitable conditions after the action. Prioritize recoverable energy, intervention cost and operational risk, rather than choosing the largest modeled percentage automatically. A design change can trade lower losses against fewer modules or higher cost.
Worked diagnostic example
These numbers are an illustrative worksheet, not a customer case or recommended acceptance threshold.
Assume a month has a historical forecast of 100 MWh at the delivery meter. A model using that month’s weather and the same declared boundary expects 90 MWh. The validated meter records 86 MWh.
| Comparison | Calculation | Result |
|---|---|---|
| Difference from historical forecast | 100 − 86 | 14 MWh |
| Difference from corresponding-weather expectation | 90 − 86 | 4 MWh |
| Measured/expected energy ratio | 86 / 90 | 0.9556, rounded |
| Residual deficit relative to expectation | (90 − 86) / 90 | 4.44%, rounded |
The first comparison mixes reference-weather variation with other differences. The second narrows the investigation but does not prove that all 4 MWh is a repairable fault. Configuration, sensor uncertainty, model limitations and operational events still need checking.
If an outage assessment estimates missing energy, state its method and whether outages were already represented in the expectation. Do not subtract the same event twice.
How to read a simulation loss diagram
Keep the original useful categories: shading, temperature, soiling, mismatch, wiring, inverter conversion, clipping and downtime. Their relative importance is site- and method-dependent; there is no universal ranking or total-loss allowance.
PVsyst’s loss-diagram documentation explains that displayed sequential loss percentages refer to the preceding energy quantity. These percentages cannot simply be summed. It also notes limitations in separating interacting contributions. Read the stage energy values, units and model definitions before interpreting a category as an independently recoverable amount.
For the detailed engineering-report context, use the PVsyst report review guide. For shading mechanisms, use shading analysis.
Report the result without an unsupported guarantee
A useful report contains the boundary/register, reference model, data-quality checks, observed patterns, confirmed causes, unresolved hypotheses and action plan. State whether a number is modeled, measured or inferred.
Performance ratio can help summarize performance, but one value does not identify a fault. Do not declare every system below a universal PR or percentage-difference threshold defective.
SurgePV publishes this definition and sells solar software. If you are evaluating its generation and financial workflow, ask for the relevant inputs, loss definitions, exported outputs and revision controls for your project. This page does not establish monitoring diagnostics, simulation equivalence or automatic lender acceptance. Use a requirements demonstration to verify the required workflow.
Questions that arise during a loss investigation
Is lower output than the proposal proof of a fault?
No. Check the proposal’s weather basis, reporting period and energy boundary, then compare valid measurements with a suitable expectation. A residual shortfall is evidence to investigate, not a diagnosis.
Can a loss diagram tell me exactly what cleaning will recover?
Not by itself. A modeled soiling assumption is not a measured cleaning benefit. Support an intervention estimate with site observations and a comparison that accounts for weather, operating status and other changes.
Is degradation the same as a bad month?
No. A single month can reflect weather, outages, data quality or model error. Establishing a long-term performance trend requires an appropriate multi-period method and documented corrections; it is a separate task from attributing one month’s deficit.
Sources reviewed September 30, 2026. Examples are illustrative calculations; no first-hand field investigation is claimed.
Related Glossary Terms
About the Contributors
CEO & Co-Founder · SurgePV
Keyur Rakholiya is identified by SurgePV as its CEO and a company co-founder. His SurgePV author page lists only role information that can be tied to the public profile below; credentials, project totals, testing claims, media appearances, and speaking engagements are not asserted without retained evidence.
Editorial contributor · SurgePV
Rainer Neumann is credited as an editorial contributor on SurgePV content. This profile does not assert engineering credentials, project totals, software-testing experience, education, speaking engagements, or media citations because independent verification evidence is not retained in the publication record.
