Two-year-old 4 kWp system producing 'half of what it did'
A 4 kWp system commissioned two years ago is reported by the customer as 'making about half what it used to'. The array is two strings of 5 x 400 W modules on a west-facing roof at 35 degrees, paralleled through a Y-branch connector pair onto a single MPPT. The visit is in late June at 12:30 on a clear day, plane-of-array irradiance 780 W/m2, modules clean, no shade at that time. The inverter logs 1.6 kW and shows no fault. With the inverter and both strings isolated at the DC isolator: string 1 - Voc 196 V, Isc 8.9 A, insulation resistance >200 MOhm; string 2 - Voc 157 V, Isc 4.0 A, insulation resistance 0.6 MOhm (tested at 1000 V).
1. Interpret these readings and state what you conclude and what you would do next.
Read the worked answer
Is the output low? 4 kWp at 780 W/m2 POA, with modules around 55 degC (roughly 12 % thermal loss) and inverter efficiency around 97 %, should give approximately 4 x 0.78 x 0.88 x 0.97 = 2.7 kW. The inverter is logging 1.6 kW, about 60 % of expected - so yes, there is a genuine fault, not a customer perception problem.
String 1 is healthy: Voc 196 V is about 5 x 39 V, consistent with five cool-to-warm modules at this irradiance, and >200 MOhm insulation is excellent.
String 2 is faulty on two counts:
- Voc 157 V is roughly 4 modules' worth, not 5. One module is not contributing - typically an open-circuited internal interconnect or a failed junction box, with the current passing through its bypass diode. On some module designs this also means a diode is now carrying full string current continuously, which is a fire risk.
- Insulation resistance 0.6 MOhm is below the 1 MOhm minimum used for PV DC circuits at the 1000 V test. That points to water ingress, a damaged cable or a degraded junction box on that string, and the reduced Isc (4.0 A against 8.9 A) is consistent with a resistive path in series.
Next steps: do not re-energise. Test each module on string 2 individually (Voc, Isc, insulation resistance to earth), inspect junction boxes and connectors for corrosion/water, check bypass diode continuity, then repair or replace the faulty module and any damaged connector. Re-test the whole string at 1000 V to >1 MOhm, re-test polarity, then re-commission. Record all readings against the original commissioning results.
2. Explain why a fault on one string has dragged the whole system down to 1.6 kW, given that the other string is healthy.
Read the worked answer
Both strings are paralleled through a Y-branch onto one MPPT, so the inverter can only choose one operating voltage for both strings simultaneously.
- String 2 can only deliver about 4 A and only has four modules' worth of voltage. The healthy string 1 is capable of around 9 A.
- The inverter's tracking algorithm settles on a compromise voltage that suits neither string. At that compromise point, string 1 is pulled off its maximum power point, so the good string is penalised by the bad one.
- The result is not 'the array loses one module' - it is a compound loss: the missing module, the current limitation, and the mismatch loss on the healthy string, plus the inverter possibly not tracking properly at all.
This is the classic argument against paralleling dissimilar strings on a single MPPT. It also explains why the customer saw a gradual decline rather than a step change - as the fault developed, the mismatch worsened.
The diagnostic lesson: measure string currents and voltages separately before blaming the inverter, and treat a healthy-looking string in a shared-MPPT array as a suspect until proven otherwise.
3. What should have been done differently at the design, commissioning and handover stages that would have prevented or quickly diagnosed this?
Read the worked answer
Design:
- Two strings of five via a Y-branch onto one MPPT is a compromise. A single string of ten would have needed checking against the cold-temperature Voc (10 x 41.2 x 1.10 = 453 V - within a 600 V inverter), removing the parallel mismatch risk altogether; alternatively a twin-MPPT inverter would have isolated the fault to one input.
- If paralleling is unavoidable, the strings must be identical in length, orientation, tilt and shading, and the design should say so.
Commissioning:
- Record string Voc, Isc and insulation resistance separately for each string and keep them as the benchmark. With those figures on file, this fault would have been spotted in minutes rather than requiring a full diagnostic visit.
- An insulation resistance test of 0.6 MOhm would have been visible at commissioning if the test had been done properly, with the inverter and any SPD disconnected, at 1000 V, and each string tested individually.
Handover/maintenance:
- Set up monitoring with a yield alert so a 40 % drop is flagged automatically rather than discovered two years later by chance.
- Provide the handover pack with the array layout, string configuration and commissioning results, and agree a first-year inspection.
- Note for the report: the bypass diode now carrying full string current is a fire risk and a reason to isolate immediately rather than wait for a convenient visit.