Analyzing Solar Plant Efficiency with Inverter Logs

Find lost kWh using SUN2000 inverter logs: IV curve analysis, string degradation detection, energy benchmarking, and FusionSolar AI comparison.

When a solar plant generates 5 to 10% less energy than its solar calculator forecast, the shortfall is rarely a mystery — the inverter has already recorded the cause. The Huawei SUN2000 logs IV curves, per-string voltage histories, and generation totals that together pinpoint whether the problem is shading, dirty panels, a failing connector, or simple degradation. This guide explains how to read each log to find the source of lost generation.

Finding lost kWh

Start with the gap between forecast and actual generation. Pull perfmg_data and compare the accumulated kWh figures against your PVsyst or solar calculator estimate for the same period. A deviation of more than 10% over a month is a reliable indicator that something beyond normal degradation is affecting output.

Once you have confirmed the gap exists, use the logs below to identify the cause.

IV curve analysis with sun_inpt_rec

The sun_inpt_rec log stores the IV curve (current-voltage characteristic) for each string. The shape of the curve tells you what is wrong before you drive to the site.

• Steps on the curve indicate partial shading. Each step corresponds to a bypass diode activating because one or more cells in that section are shaded. Common culprits are a chimney casting a shadow across a row of panels, an overhanging tree branch, or a neighboring panel (for example, on a south-east facing roof where morning shadows from a ridge or parapet cross adjacent rows).
• A flat but low curve — where Isc is uniformly reduced but the shape is otherwise normal — indicates that all cells are receiving proportionally less light. The most common cause is soiling: dust, bird droppings, or pollen film on the panel surface. These panels need cleaning, not replacement.

Run IV curve scans on a clear day between 10:00 and 14:00 with irradiance at or above 600 W/m². Scans taken in low light or partial cloud are harder to interpret because irradiance variation can mimic shading signatures.

String degradation detection with his_inv_rd

The his_inv_rd log records voltage and current for each string at 5-minute intervals. This makes it possible to compare strings that are receiving the same irradiance and identify outliers.

If one string’s voltage fluctuates 10 to 20 V more than adjacent strings under the same solar conditions, the cause is almost always a burning contact — either in an MC4 connector or inside a panel junction box. A contact with elevated resistance heats up under load, causing its voltage drop to vary with current (and therefore with irradiance), which produces the observed fluctuation.

A burning contact in an MC4 connector or junction box is a direct fire risk. If his_inv_rd shows one string’s voltage oscillating 10–20 V more than its neighbors under identical sun conditions, treat it as an urgent finding. Inspect the string connectors and junction boxes physically at the earliest opportunity — do not defer this to the next scheduled visit.

Energy benchmarking with perfmg_data

Use perfmg_data to compare actual cumulative generation against your PVsyst forecast at daily, monthly, and annual granularity. A deviation of more than 10% sustained over several weeks — after accounting for genuinely cloudy weather periods — is the threshold at which panel cleaning typically becomes cost-effective.

Track the ratio of actual to forecast generation month by month and plot it over time. A gradual downward trend over 12 to 18 months usually indicates soiling accumulation or early-stage PID (Potential-Induced Degradation), while a sudden step-change in the ratio points to a new shading obstruction or a connector failure.

FusionSolar AI benchmarking

If your inverter is connected to FusionSolar, the platform compares your plant’s performance against similar plants in the same geographic region — installations with comparable capacity, tilt angle, and azimuth that experience the same weather.

This benchmark eliminates the uncertainty of using a static PVsyst model, which cannot account for regional weather anomalies. If similar plants in your area are generating at 100% of their modelled output and yours is at 90%, the platform will flag your plant and prompt you to investigate. This type of comparison is especially useful for detecting gradual soiling or degradation that would be difficult to notice when looking at a single plant’s data in isolation.