System Size + general observations about quotes

[kinetoplast1973]

Here is a short synposis of my experience so far. I am using about 600KWH/ month. I have promised by 6 different contractors that my system is fairly simple with no shading issues or panel upgrades. But all 6 do not seem to agree on the size of system which ranges from 3.75 KW to 4.85 (100% coverage). Based on the my calculation + some help, I tend to believe that the system needs to be about 4 KWH in size. The price seems to range from $3.69-$4.75/watt. The lower $/watt vendors have much bigger system +20 year power production guarantee. This leads to much higher amount than more appropriately sized system making it unfeasbile. Just wondering, if anyone else has experienced this. Based on some research, I am looking for 13-14 LG305 panels + enphase microinvertors.

PS: I have read threads with $ 3.5/watt installation in SD but I am unable to find someone with similar pricing.

[sensij]

In San Diego, to estimate production on an unshaded array I would use PVWatts with Module Type = "Premium", Array Type = "Fixed (roof mount)", System Losses = "8%" and your azimuth and tilt.

Are they including a panel upgrade in the price? For ~4 kw, it seems unlikely to be necessary. Edit: nm, I see you mentioned this.

I can PM some 2nd hand information on lower cost installers, but hopefully members who have actually gotten that pricing will see this and jump in.

My experience was that most installers are full of #$%^.

[kinetoplast1973]

In San Diego, to estimate production on an unshaded array I would use PVWatts with Module Type = "Premium", Array Type = "Fixed (roof mount)", System Losses = "8%" and your azimuth and tilt.

Are they including a panel upgrade in the price? For ~4 kw, it seems unlikely to be necessary. Edit: nm, I see you mentioned this.

I can PM some 2nd hand information on lower cost installers, but hopefully members who have actually gotten that pricing will see this and jump in.

My experience was that most installers are full of #$%^.

Thanks. I will appreciate any recommendations.

[Naptown]

In San Diego, to estimate production on an unshaded array I would use PVWatts with Module Type = "Premium", Array Type = "Fixed (roof mount)", System Losses = "8%" and your azimuth and tilt.

Are they including a panel upgrade in the price? For ~4 kw, it seems unlikely to be necessary. Edit: nm, I see you mentioned this.

I can PM some 2nd hand information on lower cost installers, but hopefully members who have actually gotten that pricing will see this and jump in.

My experience was that most installers are full of #$%^.

I would think 8% losses would have the effect of oversizing the system
How did you arrive at that derate factor?

[kinetoplast1973]

I would think 8% losses would have the effect of oversizing the system
How did you arrive at that derate factor?

This is too technical for me. I just inputed the numbers in couple of DIY solar panel websites to get estimate of size:
To be super accurate:

My roof is:Roof Details

Roof Pitch 23° / 23°
Roof Azimuth 189° / 279°
Annual Shading 99% / 91%

[Naptown]

You have a south and northwest orientation
You will need to calculate these seperatly to arrive at overall system output.

[sensij]

I would think 8% losses would have the effect of oversizing the system
How did you arrive at that derate factor?

8% might be a little conservative, 6% is probably a better representation of my own system, but I have a hard time pushing the derate that low on a system I know nothing about.

I compare the actual power (and energy) output of my system on a clear day to what PVwatts models. I have a revenue grade production meter, and have cross-checked the numbers by other means and am confident that what it is reporting is accurate to within 0.5%. We've had lots of clear weather here lately, so I can create the chart below looking at yesterday, for example:

PVWatts1.gif

If you'd rather look at energy, for clear skies recently or around this week (PVWatts):

Actual energy recently: 20.3 - 21.0 kWh
PVWatts modeled, 4% loss: 21.1 - 21.7 kWh
PVWatts modeled, 6% loss: 20.6 - 21.2 kWh
PVWatts modeled, 8% loss: 20.2 - 20.8 kWh

Something between 6-8% looks like it represents my system reasonably well. I've only had it running for a couple months, and with more data I can improve my confidence in this recommendation, but I've found that the same derate describes other recent San Diego systems (using LG panels, too) using data from PVOutput at other times of year.

[sensij]

My roof is:Roof Details

Roof Pitch 23° / 23°
Roof Azimuth 189° / 279°
Annual Shading 99% / 91%

Do you intend to put panels on both faces of roof? The panels on the 189 deg roof will definitely produce the most, and will give you the best bang for your buck.

[kinetoplast1973]

Do you intend to put panels on both faces of roof? The panels on the 189 deg roof will definitely produce the most, and will give you the best bang for your buck.

We should be able install 11 panels on south side and remainder( 2-3) on west

[Naptown]

I seem to have made a mistake regarding losses in PVWatts
Using the 8% losses in the derate section would undersize the system.
Use the default 14% this takes all into account.
Comparing predicted vs actual and adjusting from that will most likely throw off the result.
Predicted is based on 20 years of weather data. Your difference is set at actual which could be off even more than the 8%.
I assume that is how you came up with that number.

[kinetoplast1973]

I seem to have made a mistake regarding losses in PVWatts
Using the 8% losses in the derate section would undersize the system.
Use the default 14% this takes all into account.
Comparing predicted vs actual and adjusting from that will most likely throw off the result.
Predicted is based on 20 years of weather data. Your difference is set at actual which could be off even more than the 8%.
I assume that is how you came up with that number.

Actually, my system is based on very simplistic calculation. I took my average usage and inputted in various solar power websites for a ideal system size. Since I have microinvertors, I can increase the size later on or pay the difference in bills.

[sensij]

I seem to have made a mistake regarding losses in PVWatts
Using the 8% losses in the derate section would undersize the system.
Use the default 14% this takes all into account.
Comparing predicted vs actual and adjusting from that will most likely throw off the result.
Predicted is based on 20 years of weather data. Your difference is set at actual which could be off even more than the 8%.
I assume that is how you came up with that number.

Yes, weather varies, and that is addressed explicitly in the PVWatts documentation:

Because weather patterns vary from year to year, the values in the tables are better indicators of long-term performance than of performance for a specific month or year. Photovoltaic system performance is largely proportional to the amount of solar radiation received, which may vary from the long-term average by ±30% for monthly values and ±10% for annual values.

However, by using *clear sky* data only, much of the weather variance is set aside and comparisons between the model and actual values take on more significance.

For the sake of the OP and others, here is a short primer:

PVWatts doesn't predict system output; it is a toolkit of models that relates PV generated AC power to a set of provided inputs, and is based on published, peer reviewed science. (Tangentially, I have been involved in a collaboration that recreates PVWatts and some elements of SAM in python, and learned a lot about the models used by NREL in that endeavor). The most critical of those inputs are:

1) Position of the sun - well known, NREL solar position model is extremely accurate.
2) Array orientation - well known, actual values are provided by user, not modeled.
3) Weather - documented well, but has high variance.

The role of weather affects output in a number of ways, but the two most important are:
1) Solar Irradiance, and the proportion of irradiance that is direct from the sun as opposed to diffused or reflected.
2) Temperature, not just the ambient temperature, but also wind speed, which affects the operating temperature of the array.

PVWatts also assumes the array is crystalline silicon and the model converting irradiance to power is based on that, with a small allowance for reflectivity and thermal coefficient variance handled by the "Module Type" selection. Much of the equipment and installation variance can just roll up into the "system losses".

On a *clear day*, the variance of solar irradiance is much less than it is when clouds are also considered, and the difference between the "typical" values used by PVWatts and the actual values on a day of interest becomes small.

Likewise, the temperature and windspeed in the "Typical Meteorological Year" (TMY) weather file used by PVwatts can be compared to actual temperature and windspeed, and as long as there is reasonable agreement, this will not be a big source of error between the model power output and actual power output.

The biggest source of variance between PVWatts and actual production on monthly and yearly time frames is that PVWatts does not know how many cloudy days occur each month (or year), and it doesn't know if it is unusually cold (or hot) over the time period. However, by looking in the TMY file and identifying clear days at around the same time of year with temperature characteristics similar to actual values, the model provides very good agreement with reality, with loss factor remaining as the biggest unknown. That is the technique I used in my post above to estimate the loss factor for my system.

Summing all that up... I can't tell you with high accuracy how much energy I will produce in June 2015. What I can say is that if the data representing June 25 in the TMY is based on a clear sky and the actual weather on June 25, 2015 is clear, they will agree reasonably well, as long as the temperatures are about equal.

An even more rigorous way to estimate the system losses is to use NREL's System Advisor Model (SAM), and build a "live" TMY file based on actual irradiance and weather for each day. I haven't invested in a weather station that would allow me to do this, but forum member J.P.M. has and hints at employing this technique to characterize his own system. I should also give him credit for suggesting/inspiring the analysis described above that uses PVWatts in a way that provides much more insight into array performance than is obvious on first glance.

Actually, my system is based on very simplistic calculation. I took my average usage and inputted in various solar power websites for a ideal system size. Since I have microinvertors, I can increase the size later on or pay the difference in bills.

"Various solar power websites" are collectively likely to lead you astray, although there are some good ones out there. I would strongly recommend PVWatts, since the technical reference material describing how it works is freely available and based on very sound science.

Also, I would be skeptical about the idea that you can easily "increase the size later", whether you use microinverters, Solaredge optimizers, or a standard string inverter. For any of them, some planning up front should go into how much deciding expansion you might need later, so that wire of sufficient size is in place and the capacity of your solar circuits (and service panel) will accommodate your future expansion.

Especially in San Diego, you should also familiarize yourself with TOU pricing. Since your array will produce power during the part of the day that SDG&E values it most, it can be possible to offset much more of your bill than a straight kWh offset would suggest. For example, in June so far this year, I've generated 440 kWh, consumed 535 kWh, and am still tracking so far to a billing credit of about $50 that I can use to pay for consumption in winter. In other words, the value of that 440 kWh is allowing me to offset the cost of nearly 800 kWh of actual consumption.

[kinetoplast1973]

Yes, weather varies, and that is addressed explicitly in the PVWatts documentation:



However, by using *clear sky* data only, much of the weather variance is set aside and comparisons between the model and actual values take on more significance.

For the sake of the OP and others, here is a short primer:

PVWatts doesn't predict system output; it is a toolkit of models that relates PV generated AC power to a set of provided inputs, and is based on published, peer reviewed science. (Tangentially, I have been involved in a collaboration that recreates PVWatts and some elements of SAM in python, and learned a lot about the models used by NREL in that endeavor). The most critical of those inputs are:

1) Position of the sun - well known, NREL solar position model is extremely accurate.
2) Array orientation - well known, actual values are provided by user, not modeled.
3) Weather - documented well, but has high variance.

The role of weather affects output in a number of ways, but the two most important are:
1) Solar Irradiance, and the proportion of irradiance that is direct from the sun as opposed to diffused or reflected.
2) Temperature, not just the ambient temperature, but also wind speed, which affects the operating temperature of the array.

PVWatts also assumes the array is crystalline silicon and the model converting irradiance to power is based on that, with a small allowance for reflectivity and thermal coefficient variance handled by the "Module Type" selection. Much of the equipment and installation variance can just roll up into the "system losses".

On a *clear day*, the variance of solar irradiance is much less than it is when clouds are also considered, and the difference between the "typical" values used by PVWatts and the actual values on a day of interest becomes small.

Likewise, the temperature and windspeed in the "Typical Meteorological Year" (TMY) weather file used by PVwatts can be compared to actual temperature and windspeed, and as long as there is reasonable agreement, this will not be a big source of error between the model power output and actual power output.

The biggest source of variance between PVWatts and actual production on monthly and yearly time frames is that PVWatts does not know how many cloudy days occur each month (or year), and it doesn't know if it is unusually cold (or hot) over the time period. However, by looking in the TMY file and identifying clear days at around the same time of year with temperature characteristics similar to actual values, the model provides very good agreement with reality, with loss factor remaining as the biggest unknown. That is the technique I used in my post above to estimate the loss factor for my system.

Summing all that up... I can't tell you with high accuracy how much energy I will produce in June 2015. What I can say is that if the data representing June 25 in the TMY is based on a clear sky and the actual weather on June 25, 2015 is clear, they will agree reasonably well, as long as the temperatures are about equal.

An even more rigorous way to estimate the system losses is to use NREL's System Advisor Model (SAM), and build a "live" TMY file based on actual irradiance and weather for each day. I haven't invested in a weather station that would allow me to do this, but forum member J.P.M. has and hints at employing this technique to characterize his own system. I should also give him credit for suggesting/inspiring the analysis described above that uses PVWatts in a way that provides much more insight into array performance than is obvious on first glance.






"Various solar power websites" are collectively likely to lead you astray, although there are some good ones out there. I would strongly recommend PVWatts, since the technical reference material describing how it works is freely available and based on very sound science.

Also, I would be skeptical about the idea that you can easily "increase the size later", whether you use microinverters, Solaredge optimizers, or a standard string inverter. For any of them, some planning up front should go into how much deciding expansion you might need later, so that wire of sufficient size is in place and the capacity of your solar circuits (and service panel) will accommodate your future expansion.

Especially in San Diego, you should also familiarize yourself with TOU pricing. Since your array will produce power during the part of the day that SDG&E values it most, it can be possible to offset much more of your bill than a straight kWh offset would suggest. For example, in June so far this year, I've generated 440 kWh, consumed 535 kWh, and am still tracking so far to a billing credit of about $50 that I can use to pay for consumption in winter. In other words, the value of that 440 kWh is allowing me to offset the cost of nearly 800 kWh of actual consumption.

you were right . The system was upgraded to 4.6 KW with PVWatts

[J.P.M.]

I seem to have made a mistake regarding losses in PVWatts
Using the 8% losses in the derate section would undersize the system.
Use the default 14% this takes all into account.
Comparing predicted vs actual and adjusting from that will most likely throw off the result.
Predicted is based on 20 years of weather data. Your difference is set at actual which could be off even more than the 8%.
I assume that is how you came up with that number.

Just got back in town.

On the PVWatts default value:

1.) From my experience, comparing SAM and PVWatts outputs for my zip requires the PVWatts default to be somewhere between 6% to 8% if a consistent match to match the SAM output is to be expected. A % or 2 more might be conservative. IMO, that 14 % is too conservative and leads to oversizing, at least in my area and for arrays I keep track of.

2.) Most clear day output for systems I keep track of in my HOA matches SAM clear day output for same/similar dates within a few % +/-, some more, some less.

3.) My array requires a PVWatts derate of about 6.7% to match SAM's output estimate for my array.

4.) SAM's output for my array matches well with the output of stuff I've written using TMY3 data for Miramar. My stuff will also simulate clear day output using clear day irradiance data calc'd from the HDKR model.

5.) Using clear day data and adjusting the model for actual weather data, including solar irradiance, wind vector and ambient temp. taken at 1 min. intervals with a Davis Pro II Plus located about 4 feet from my array, I'm of the opinion that my models do a reasonable job of verifying/estimating output. Since they also seem to match SAM using TMY3 data, and since PVWatts needs a derate of 6.7% or so to match my data, I'd say PVWatts needs somewhere between 6% - 8% to match what seems to be close to reality, at least for things I've measured.

6.) I do not agree with the comment that "Predicted is based on 20 years of weather data." First off, I strenuously avoid calling what a solar model may spit out as a prediction. It's a long term estimate. Second, the TMY years are 12 months of the most "typical" of each month over 15 or 30 years, "typical" meaning chosen on a weighted basis of weather parameters, and then concatenated together to form a "typical year". So, it's 12 months of data chosen from 15 or 30 years of data. Also, and I know this was not written, but TMY weather data is not an average of any weather data. Finally, FWIW, there is also weather data other than TMY2 or TMY 3 that both PVWatts and SAM can use for estimates. To my experience, all 3 get within spitting distance of one another.

7.) My array has been operating since 10/17/2013. The first year's output was 9,562 kWh for a 5.232 kW system. Since the 1st year anniv. of operation, the running 365 day average output has been as high as 9,620 kWh/yr. and as low as 9,229 kWh/yr., depending on a lot of variables including panel fouling which I've been estimating and attempting to measure for over a year now. That's a separate but somewhat related subject.

8.) SAM's estimated output annual output for my array is 9,504 kWk/yr. PVWatts estimated output using a 14% derate is 8,759 kWh/yr. PVWatts estimated output using a 6.74 % derate is 9,501 kWh/yr.