Analysis of 16 years of Solar PV data.

A friend from North London kindly allowed me to analyse the data they had collected on the performance of their solar PV installation over the last 16 years.

What an opportunity to discover how solar PV panels behave over the long term!

Let me tell you what I found:

The System

Installed in July 2006, the system consisted of 16 Sanyo PV panels, each 0.88 m x 1.32 m with a nominal peak output of 210 W. This implies the panels output was initially ~180 watts per square metre.

They were installed on two adjacent roofs with a tilt of about 30° and facing 25° East of South and with no nearby trees or shading structures on the horizon other than their neighbour’s house.

The data set consisted of roughly 700 readings of the solar generation meter, most of them taken weekly but with a couple of gaps for a few months, and few points that were clearly in error. Rather than try to be sophisticated, I simply omitted points that were obviously in error.

Click image for a larger version. The ‘cleaned up’ data set.

Annual Analysis

One of things I was most anxious to search for was evidence of a year-on-year decline. The annual results are shown below:

Click image for a larger version. Graph of the Annual Output (kWh) of a North London PV system from 2006 to 2021. The dotted line is a linear fit to the data showing a systematic year-on-year decline in output.

It’s clear that there is a systematic year-on-year decline. If we re-plot the data to express this as a percentage we can compare it with what we might expect.

Click image for a larger version. The same data as in the previous graph but expressed as a fraction of the average output over the years 2007 and 2008. The dotted line is a linear fit to the data showing a systematic year-on-year decline in output.

This decline is – sadly – inevitable, arising as I understand it from atomic defects created in the silicon cells by exposure to the UV radiation in sunlight. These defects trap electrons which would otherwise reach an external contact if the crystal had been undamaged.

A decline of 6.1% per decade (0.61% per year) is quite competitive. Older panels showed higher declines (link) and more modern cells claim better performance, but not much better.

For example a 2020 Q-Cells Duo panel (link) specifies 0.54%/year decline for up to 10 years,  i.e. 5.4% per decade.

Click image for a larger version. Extract from a Q-cells data sheet showing expected decline in panel output over 25 years.

Variability

In addition to a linear decline in output the data also shows significant year-to-year variability. I wondered whether this variability arose from the natural variability of available sunshine, or some other factor.

To check this I exploited the EU Photovoltaic Geographical Information System (a.k.a. a ‘Sunshine Database’) which allows the calculation of the output of PV cells at any point in Europe or Africa over the period 2005 to 2016.

I had previously used this database to model the year-to-year variability of sunshine in West London when I was planning a battery installation.

To see if this was the cause of the year-to-year variability I plotted two quantities on the same graph:

  • The so-called ‘residuals’ of the fit to the data in the second graph above.
  • The variability of EU-database data.

The results are shown below.

Click image for a larger version. The variability of the North London PV data and the natural variability of sunshine as retro-dicted by the EU sunshine database

It is clear that in the years for which the two datasets overlap they agree well, suggesting that the variability observed is not due to some other poorly understood factor.

Upgrade?

My North London friend had one final question. Would they avoid more carbon dioxide emissions if they upgraded to modern panels?

To answer this I made two models:

  • The first model assumed that they did not upgrade and the existing panels were used to out to 2050.
  • The second model assumed they were replaced in 2022 with panels which operated with an efficiency of around 200 W/m^2 at peak illumination. This is about 20% more than the panels currently generate.

I assumed that the new panels would embody around 2 tonnes of CO2 emissions because Q-cells suggest their latest panels embody 400 kgCO2 per kWp.

I then assumed that 50% of the generated electricity was exported and 50% used domestically. As the grid currently functions:

  • Exported electricity reduces gas-fired generation which emits 450 gCO2/kWhe.
  • Domestic use avoids consumption of grid electricity with a carbon intensity of around 220 gCO2/kWhe in 2022.

Based on these assumptions, there is small advantage to replacing the panels, but this would not be realised until 2035.

Click image for a larger version. Does it make carbon-sense to replace existing PV cells with new more efficient cells?

One can model variations of these parameters, but the basic result is not affected: the carbon advantage is marginal.

My friend would help the climate more effectively by allocating his capital expenditure to something which might have more impact on CO2 emissions, perhaps buying shares in a wind farm?

But the result that really struck me from this modelling was how great the solar panels were in the first place!

Installed in 2006 and given minimal maintenance, it looks like the existing cells will avoid almost 30 tonnes of CO2 emissions by 2050. Not many technologies can achieve results like that as easily as that.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s


%d bloggers like this: