A 50 MW plant can lose approximately R$387,000 each year across detectable fault categories. That figure shows why general claims about earlier detection are not enough. Owners assessing annual monitoring costs need a clear view of each loss category and its likely recovery. The calculations below provide a practical basis for that decision.
These figures are illustrative and reflect typical industry ranges. Actual losses vary with fault type, irradiance, PPA or market rate, and response time. Use the ranges as a starting point for your own estimate.
Loss Categories Addressed by Early Detection
Continuous monitoring does not affect every plant loss equally. Detection timing matters most in these categories:
Inverter trips and partial operation. An inverter trip stops all or part of generation in one AC circuit. If a 1 MW inverter block trips at 11:00 AM, is missed until the end of day report, and is dispatched the next morning with a 4-hour travel time, roughly 6 to 9 hours of peak-period generation are lost. At a Northeast Brazil site producing 1,800 to 2,000 full-load hours per year, a 1 MW block averages about 200 kW to 250 kW during operational hours. A 10-hour outage equals 2.0 to 2.5 MWh of lost generation. At R$320/MWh, that is R$640 to R$800 per event. With 4 to 6 inverter-trip events per year per MW installed, delayed detection can create a substantial annual loss.
String-level persistent faults. Unlike inverter trips, string faults seldom produce a SCADA alarm. A string operating at 70% of expected output because of an open-circuit module or high contact resistance can keep producing without an alarm. Monthly PR spreads the loss across its averaging window. In a 50 MW plant with 80 strings per inverter block and 12 inverter blocks, a 2% string fault rate means roughly 19 strings remain at partial output. If each affected string loses 25 kW on average, the plant has 475 kW of continuous underperformance, or roughly 0.9% of plant capacity. At 2,000 equivalent full-load hours, that equals 950 MWh of lost generation per year. At R$320/MWh, the annual loss is R$304,000.
Soiling accumulation. In Brazil's Central-West and Northeast, high dust levels can reduce output by 0.1 to 0.2% per day between cleaning events. For a 30 MW plant cleaned every 45 days, average soiling loss at the midpoint is roughly 4 to 5%. If continuous soiling monitoring shows that a 30-day cleaning cycle is more economical than a 45-day cycle at a given site, lower soiling losses can offset the added cleaning cost. The calculation depends on measuring the site's soiling rate.
Building the Illustrative ROI Case
Consider an illustrative 50 MW plant in Bahia with 2,000 equivalent full-load hours per year, a PPA rate of R$310/MWh, an O&M contractor cost of R$28/MWp/year (within the R$25 to R$35 range typical for managed plants in this region), and an average technician dispatch cost of R$2,500 per unplanned call-out.
Current reactive monitoring baseline:
- 6 inverter trips per year, 14-hour average detection-to-repair window: 84 MWh lost at 1 MW average × R$310 = R$26,040
- 2% string fault rate, 6-month average persistence before detection: roughly 700 MWh/year × R$310 = R$217,000
- 8 unplanned dispatch events per year: 8 × R$2,500 = R$20,000
- Soiling at 4% average between cleanings, 6 cleaning cycles/year: ~400 MWh loss/year × R$310 = R$124,000
Total illustrative annual loss from detectable fault categories: approximately R$387,000.
If continuous fault detection cuts the inverter trip detection-to-response window from 14 hours to 2 hours, identifies string faults within 3 days rather than 6 months, and supports better cleaning schedules through soiling monitoring, realistic recovery across these categories is 60 to 75%. For this illustrative 50 MW plant, that equals R$230,000 to R$290,000 in recoverable generation revenue per year.
For a plant this size, a monitoring system costing R$80,000 to R$150,000 per year supports a direct payback calculation. The harder question is usually not the math, but confidence in the stated detection rates.
Where Uncertainty Remains
The most uncertain input in many fault detection ROI calculations is the detection rate. Saying "we will detect string faults within 3 days" makes a specific claim about system sensitivity and plant data quality. In practice, several factors can reduce that performance.
String current data often contains substantial noise. SCADA systems with 5-minute sampling, rounding, or communication lag can hide early degradation. An anomaly that begins at 3% deviation from peer strings needs clean, low-noise data to rise above measurement uncertainty. If the SCADA historian has 5% typical data noise, reliable detection requires a 5% deviation, so the fault may already have caused losses for some time.
Changing irradiance can also distort string comparisons. Rapid fluctuations on partly cloudy days make string-to-string data noisy. Detection algorithms that overlook irradiance transients create excess false alarms, encouraging O&M teams to ignore alerts. A system with a 30% false positive rate soon loses operational credibility.
The detection claim therefore needs detail: the irradiance level, minimum fault size, false positive rate, and fault types covered. Broad statements about "early detection" do not provide enough information for a credible ROI case.
Putting Numbers to the Monitoring Decision
Start with your own plant data and estimate the loss categories above. Then ask the vendor for specific detection performance claims and supporting evidence. For a pilot, run the system against historical data containing known faults. Check whether it would have detected each fault, and how much lead time it would have provided before the fault was identified.
Continuous monitoring software does not guarantee a fixed amount of generation recovery. It provides your O&M team with better information and reduces the time between fault occurrence and response. Recovery follows when the team can act faster and target the right work. The ROI is real, but it depends on whether your O&M operation can act on the signals the system produces.
For context on the operational workflow that makes early detection actionable, see our article on reducing unplanned downtime at solar plants. For a portfolio-level view of where monitoring investment fits, see our discussion of digitizing solar asset management.