Identifying String-Level Underperformance Masked by SCADA Averaging Intervals
String-level underperformance is the silent degradation of individual PV strings that remains undetected when high-level SCADA data averages multiple sensor readings over 15-minute intervals, effectively masking transient faults.
When SCADA systems aggregate data over long intervals, they "smooth out" short-duration transients. This averaging obscures specific failures like PID, localized soiling, or intermittent connector faults. While many EPCs rely on plant-wide Performance Ratio (PR) metrics, this often creates a gap between modeled baseline PR and actual commissioning PR. If your inverter shows 98% efficiency, you might miss a string that is dragging your actual yield down by 10%.
The Math of Masking
Averaging masks low-current events by blending them with high-performing strings. If String A produces 8A and String B produces 4A (due to a fault), the SCADA log records a healthy 6A average.
The Formula: To calculate the true variance and identify outliers, compare the standard deviation of string currents during clear-sky conditions: $Variance = \sum_{i=1}^{n} \frac{(I_{i} - I_{avg})^2}{n-1}$
Numerical Example: If your site has 20 strings per combiner box and one string operates at 50% capacity, the impact on the total combiner output is only 2.5%. A 15-minute average interval makes this 2.5% loss indistinguishable from normal irradiance noise or inverter MPPT hunting behavior.
Rule of Thumb: To identify early stage string level underperformance, flag any string with a variance exceeding 5% from its peers during stable irradiance (>600W/m²), as this typically precedes hard hardware failures.
Engineering Reality Check
Engineers often need to validate these trends against historical site data to isolate performance degradation from environmental variables. You can test your calculations and refine your performance models using the SolarMetrix performance simulator at solarmetrix.app/tool.
5 Causes of Hidden String Underperformance
- PID (Potential Induced Degradation): Causes a gradual voltage drop across modules, often disguised as typical thermal aging.
- Micro-cracks: Invisible to the eye, these create high-resistance paths that fluctuate with temperature.
- Intermittent MC4 Connections: High resistance at the junction creates heat, triggering intermittent drops that vanish before the next SCADA log.
- Localized Soiling: Bird droppings or shade patterns that occur outside of your standard modeled irradiance profiles.
- String-Level Fuse Degradation: Fuses that haven't blown but have developed internal resistance, causing a constant, low-level power bottleneck.
Diagnostics Strategy
Stop trusting your SCADA dashboard. Start analyzing current data at the highest possible sampling frequency to combat SCADA data granularity masking short-duration inverter trips.
- Audit for Delta: Run a correlation between string current and the site pyranometer during steady-state irradiance to ensure plant performance ratio distortion isn't caused by incorrect plane-of-array irradiance measurement.
- Identify Outliers: Filter for strings consistently 3-5% below the mean.
- Check Thermal Profiles: Use IR drone inspections to correlate low-current strings with localized heating.
If you don't look at the raw data, you aren't managing the asset; you're just watching the meter.
FAQs
How do I detect string underperformance if I only have 15-minute SCADA data? You cannot effectively detect intermittent faults with 15-minute data. You must request high-resolution (1-minute or less) raw data from the inverter’s internal logger. Compare individual string current against the average of all strings connected to the same MPPT. Consistent outliers, even if small, indicate a physical fault.
What is the minimum performance threshold to trigger a site inspection? Any string showing a consistent deviation of >3% from the average current of its sibling strings under stable irradiance (above 600W/m²) should trigger an inspection. Do not wait for the inverter to throw an alarm; string-level failures rarely trigger inverter-level error codes until the component has completely failed.
Does increasing the SCADA sampling rate impact data storage costs? High-frequency data creates massive datasets. Instead of storing 1-second data permanently, store only the calculated deviation statistics or trigger high-resolution logging only when irradiance is stable (clear-sky periods). This reduces storage overhead while capturing the granular data necessary to identify masked underperformance before it impacts your project's bottom line.