In this document:
This article explains why accurate PV system configuration is critical for reliable PV power output forecasts, and how Solargis uses the PV configuration estimator tool to verify and correct customer-provided PV configuration data. Three real-life case studies demonstrate the impact of PV configuration errors on forecast accuracy and the improvements achieved by deriving effective PV configuration parameters from measured PV power output data.
Overview
Accurate PV power output forecasting depends on two key inputs: solar irradiance data from weather prediction models and correct PV system configuration metadata. In practice, PV configuration metadata provided by customers — referred to as the declared configuration — is often incomplete or incorrect. Even when the irradiance forecast is fully accurate, incorrect PV configuration parameters introduce systematic deviations between forecast and reference data, leading to financial and logistical problems for grid operators and PV asset owners.
To address this, Solargis uses the PV configuration estimator tool to verify declared PV configuration parameters against measured PV power output data. The tool derives an effective PV configuration — the set of parameters that best represents the actual behavior of the PV system — by running an optimization loop over the reference dataset of measured PV power output. The effective PV configuration minimizes systematic deviations and improves forecast accuracy.
Key parameters verified and adjusted by the tool include:
Installed capacity [kW]
Azimuth (panel orientation, North–South convention) [°]
Tilt angle [°]
Panel geometry type
Column / row spacing
Future development of the tool is planned to extend verification to additional parameters including AC power limitation, PV system losses and PV configuration parameters specific to PV systems with tracking mechanisms.
Why correct PV configuration matters
PV power output is derived from Global Horizontal Irradiance (GHI) using a transformation chain that depends on PV system configuration metadata. The transformation process is summarized as:
Each step in this chain relies on PV configuration parameters such as azimuth, tilt, installed capacity and more. An error in any of these parameters introduces a systematic bias in the resulting PVOUT forecast — independent of the accuracy of the upstream irradiance data.
Note: Systematic deviations caused by incorrect PV configuration cannot be corrected by improving the irradiance forecast alone. The PV configuration itself must be verified.
The PV configuration estimator tool resolves this by comparing satellite-based PVOUT (generated using the declared PV configuration) against reference PVOUT measured at the site. Discrepancies trigger an optimization process that produces the effective PV configuration.
Forecast accuracy metric
Forecast accuracy in all case studies is evaluated using the normalized Mean Absolute Error (nMAE), expressed as a percentage of the installed capacity:
where:
N = number of data points in the evaluation period
PVOUT_forecast = forecast PV power output [kW]
PVOUT_reference = measured (reference) PV power output [kW]
P_installed = installed capacity [kW]
Results are presented on both monthly and yearly basis, covering a 12-month reference period (August 2022 – July 2023).
Case studies
The following three case studies illustrate how the PV configuration estimator tool improves forecast accuracy across different types of PV configuration errors.
Case study 1 — Incorrect installed capacity
The first demonstration site is a building-integrated PV system. The declared installed capacity was 1,546 kW. Initial visual comparison of forecast and reference PV power output profiles revealed a large and consistent discrepancy, indicating a significant systematic error in the PV configuration.
Figure 1: Sequence of PV power output daily profiles — forecasts based on the declared PV configuration vs. reference.
The PV configuration estimator tool was applied to one year of reference data (August 2022 – July 2023). The table below summarizes the differences between the declared and effective PV configuration parameters.
System ID | Declared installed capacity [kW] | Declared azimuth [°] (North–South) | Declared tilt [°] | Declared geometry | Estimated installed capacity [kW] | Estimated azimuth [°] (North–South) | Estimated tilt [°] | Estimated geometry |
|---|---|---|---|---|---|---|---|---|
1 | 1,546 | 180 | 20 | FIXEDONEANGLE | 820 | 172 | 15 | FIXEDONEANGLE |
Table 1: Comparison of the declared and estimated (effective) PV configuration parameters.
The most significant correction was to installed capacity, which was reduced from 1,546 kW to 820 kW — a reduction of approximately 47%. The tool also made minor adjustments to azimuth and tilt.
Important: A declared installed capacity almost twice the effective value is not an exceptional case. It reflects a common real-world scenario where nameplate capacity does not correspond to the operationally active capacity of the PV system.
Forecasts were then generated using the effective PV configuration and evaluated against the reference data.
Results
The effective PV configuration substantially improved forecast accuracy across the entire evaluation period. The primary driver of error in the declared PV configuration was the significantly overstated installed capacity — nearly double the effective value — which introduced a large and consistent overestimation of PV power output. Correcting this parameter reduced nMAE in every month of the 12-month reference period.
Figure 2: Comparison of monthly normalized Mean Absolute Errors (nMAE) — declared vs. effective PV configuration.
Figure 3: Comparison of yearly normalized Mean Absolute Errors (nMAE) — declared vs. effective PV configuration.
The effective PV configuration significantly reduced nMAE across all 12 months. The visual comparison of forecast and reference daily profiles confirms the improvement.
Figure 4: Sequence of PV power output daily profiles including effective forecasts.
Case study 2 — Incorrect azimuth
The second demonstration site is a fixed-angle PV system. In this case, the declared and estimated installed capacities are identical at 600 kW. However, the azimuth shows a large discrepancy: the declared value is 135° while the effective value estimated by the tool is 180°. The tilt was also adjusted from 15° to 5°.
The initial comparison of forecast and reference profiles reveals systematic deviations caused by the incorrect orientation and tilt.
Figure 5: Sequence of PV power output daily profiles — forecasts based on the declared PV configuration vs. reference.
System ID | Declared installed capacity [kW] | Declared azimuth [°] (North–South) | Declared tilt [°] | Declared geometry | Estimated installed capacity [kW] | Estimated azimuth [°] (North–South) | Estimated tilt [°] | Estimated geometry |
|---|---|---|---|---|---|---|---|---|
2 | 600 | 135 | 15 | FIXEDONEANGLE | 600 | 180 | 5 | FIXEDONEANGLE |
Table 2: Comparison of the declared and estimated (effective) PV configuration parameters.
Results
The effective configuration improved forecast accuracy consistently across all 12 months. In this case, the installed capacity was correct, but the incorrect azimuth (135° declared vs. 180° effective) caused a systematic shift in the daily power output profile shape. Correcting the orientation and tilt brought the forecast into closer alignment with the reference data, though the overall improvement is less pronounced than in Case study 1.
Figure 6: Comparison of monthly normalized Mean Absolute Errors (nMAE) — declared vs. effective PV configuration.
Figure 7: Comparison of yearly normalized Mean Absolute Errors (nMAE) — declared vs. effective PV configuration.
The improvement in forecast accuracy is consistent across all 12 months, though less dramatic than in Case study 1. The visual check of daily profiles confirms the improved alignment between forecast and reference data.
Figure 8: Sequence of PV power output daily profiles including effective forecasts.
Case study 3 — Small but impactful PV configuration errors
The third demonstration site is a fixed-angle PV system. In this case, the differences between declared and effective PV configuration parameters are relatively small, as summarized in the table below.
System ID | Declared installed capacity [kW] | Declared azimuth [°] (North–South) | Declared tilt [°] | Declared geometry | Estimated installed capacity [kW] | Estimated azimuth [°] (North–South) | Estimated tilt [°] | Estimated geometry |
|---|---|---|---|---|---|---|---|---|
3 | 115 | 180 | 7 | FIXEDONEANGLE | 112 | 170 | 5 | FIXEDONEANGLE |
Table 3: Comparison of the declared and estimated (effective) PV configuration parameters.
The declared PV configuration was corrected to an effective PV configuration. Despite the modest differences, the declared PV configuration produces a visible systematic bias when compared against reference data.
Figure 9: Sequence of PV power output daily profiles — forecasts based on the declared PV configuration vs. reference.
Results
Despite the small magnitude of the PV configuration corrections, forecast accuracy improved in each of the 12 individual months. This result highlights that even minor discrepancies in PV configuration parameters — in this case less than 4 kW in capacity, 10° in azimuth, and 2° in tilt — can introduce a measurable systematic bias that the PV configuration estimator tool is able to detect and correct.
Figure 10: Comparison of monthly normalized Mean Absolute Errors (nMAE) — declared vs. effective PV configuration.
Figure 11: Comparison of yearly normalized Mean Absolute Errors (nMAE) — declared vs. effective PV configuration.
Forecast accuracy improved in each of the 12 individual months. This result demonstrates that even small corrections to PV configuration parameters can lead to a meaningful improvement in forecast accuracy.
Figure 12: Sequence of PV power output daily profiles including effective forecasts.
Summary of results
The three case studies cover a range of PV configuration error types and magnitudes. In all cases, the PV configuration estimator tool improved forecast accuracy on both a monthly and yearly basis. The table below summarizes the primary corrections made in each case and the nature of the accuracy improvement.
Case / PV system ID | Primary correction | Influence on accuracy |
|---|---|---|
1 | Capacity: 1,546 → 820 kW | Significant improvement across all months |
2 | Azimuth: 135° → 180°; Tilt: 15° → 5° | Consistent improvement across all months |
3 | Capacity: 116 → 112 kW; Azimuth: 180° → 170°; Tilt: 7° → 5° | Measurable improvement across all months |
Tip: Measured PV power output (customer data) is always helpful when managing forecast accuracy. Even a single year of reference data from a customer is sufficient for the PV configuration estimator tool to derive a reliable effective PV configuration.
Planned improvements
The current version of the PV configuration estimator tool verifies installed capacity, azimuth, tilt, geometry type and column / row spacing. Future development will extend verification to:
AC power limitation
PV system losses
PV configuration parameters specific to PV systems with tracking mechanisms
Usage in Solargis platform
The PV configuration estimator tool is used as part of the Solargis Forecast data service workflow.