Solar PV Performance Analysis: 3.5. Case Study

We've been through some of the major sources of underperformance of a solar PV site during the previous posts, but each site has its own particularities and potential sources of underperformance based on design or local site conditions.

Today we'll take a concrete real-life solar PV plant and analyse it to assess the performance and corrective actions. This example is a rooftop solar installation with zero grid export, but the same principles apply to any type of setup (grid feeding or ground mounted or floating).

Here is an overhead layout of the site, there are two rooftops connected to two different meters:

Case study site layout

For this case study we assume that there is no inverter or string level unavailability, therefore any low PR is caused by underperformance of equipment compared to PV Syst or other model simulation.

When looking at the inverter PRs, we see the following patterns:

1. Generally lower PRs for roof 1 inverters vs roof 2 inverters

2. Some inverters with lower PRs and late starting in the mornings

3. Some inverters with lower PRs and early finishing in the afternoons

4. Dips on all inverter production at midday

To start with, even if you have already physically visited the site, it’s still good to do a virtual tour to understand anything that could impact performance.

The easiest way to do a virtual tour is using Google Maps or Google Earth, we take the GPS coordinates of the site and place ourselves at the street view. In this case we stand at the blue star indicated on the layout above, looking in the direction of the blue arrow:

Street view of the site showing jackroofs

From this we can see the roof has a gentle north/south orientation but has jack roofs (raised roof sections) at the apex which will cause shading to surrounding modules.

If we now focus on roof 1 we can plot out the areas that will be shaded at different times of the day. Then using the inverter & string layout (as-built) of the site, we can work out which inverters/string will be shaded at what time of day:

Indicating the impact of jackroof shading

From that we can compare the performance of the non-shaded inverters to the performance of the shaded inverters to calculate the estimated shading losses. These losses should then be compared to the PV Syst / financial model assumptions. In our case the shading was not factored into the budget so this shading is an underperformance.

Monitoring data showing shading

In the above monitoring chart we see shading on two inverters in the morning and shading on two inverters in the afternoon, this is caused by the jackroofs highlighted in the layout above.

Shading is difficult to remediate once the modules are installed, but rewiring strings (e.g. having the string linear close to the jackroof) could be an option to help reduce the impact of the shading.

Generally modules should be wired linearly and parallel to shading objects, to minimise the impact of the shading.

Also the feedback loop needs to happen to the engineering team to ensure better simulation of losses during site design (and specifically the engineers doing the PV Syst simulation should have visited the site beforehand!).

So going back to our conclusions from the inverter PR analysis, we found the following: 

1. Lower PRs for all roof 1 inverters vs roof 2 inverters

2. Some inverters with lower PRs in the mornings

3. Some inverters with lower PRs in the afternoons

4. Dips on all inverter (roof 1 & 2) production at midday

This shading helps to explain points 2 and 3. But it doesn’t help to explain point 1, as not all roof 1 inverters are affected by shading and therefore they shouldn’t all be underperforming vs roof 2 inverters. It also doesn't help to explain point 4.

For point 1, when looking at the layout, we notice that roof 1 is further away from its injection point (meter 1) vs roof 2. Roof 1 modules are around 600m away from meter 1. As we know that longer DC cable leads to higher resistance and higher voltage drop along the cable, then longer DC cable runs will lead to lower performance.

In our case the same 6mm2 DC cable has been used for all strings, and with the 600m distance this leads to approximately 1.5% voltage drop and impact on performance. This helps to explain why PR is generally lower for meter 1 inverters than meter 2.

Again this wasn’t captured during the design phase so this is real “underperformance” which is now costly to remedy, because it would require a complete re-string using e.g. 10mm2 DC cable. The gains in energy would surely not justify the cost of the exercise (to be confirmed with a cost/benefit analysis). So this is a feedback loop towards engineering on future projects to factor this in when doing the simulation.

So that leaves point 4. When we check the monitoring data there is a big dip in output at midday, when we'd expect output to be increasing:

Drop in all inverter output around midday

This isn't related to irradiance as the irradiance curve is a perfect bell shape (not shown here). Also we know that this isn't clipping as the inverter output would flatline rather than drop (see article 3.3 for more details on clipping).

What we do know is that this is a zero export site, and at lunchtime employees will go to lunch and factory production will slow, leading to a reduced demand for electricity. So this drop in output is a curtailment loss. Based on the PPA contract with the client, this loss may be claimable as part of the client's annual minimal consumption commitment.

For this site we have reviewed some major sources of undeperformance:

1. Lower PRs for all roof 1 inverters vs roof 2 inverters --> because of 1.5% voltage drop caused by longer DC cable route

2. Some inverters with lower PRs in the mornings --> because of jackroof shading

3. Some inverters with lower PRs in the afternoons --> because of jackroof shading

4. Dips on all inverters (roof 1 & 2) production at midday --> because of factory curtailment when workers go to lunch