Announcement

Collapse
No announcement yet.

PVWatts - How close are you?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • PVWatts - How close are you?


    When I designed my solar system, I used NREL's PV Watts calculator to determine sizing. My solar system is on its 5th full day of energy production and I'm watching what was calculated (via PV Watts) vs what I'm really getting..
    Obviously, five days of data is not even remotely close enough to make any determinations but there are certain things I can decipher. For instance, when I scan the PV Watts hourly data spreadsheet to see what my panels should be producing under full sun at noon, I can scroll the entire month and see what the highest PV Watts estimated numbers are and then compare my panels to those numbers when they're under full sun at noon. Again, not super accurate as I'm sure temperature and humidity in the air will make a difference..

    I'm curious as to the results others have when it comes to months or years of data.. Did you come in over the PV Watts estimations or under.... and by how much?


  • #2
    Originally posted by Murby View Post
    When I designed my solar system, I used NREL's PV Watts calculator to determine sizing. My solar system is on its 5th full day of energy production and I'm watching what was calculated (via PV Watts) vs what I'm really getting..
    Obviously, five days of data is not even remotely close enough to make any determinations but there are certain things I can decipher. For instance, when I scan the PV Watts hourly data spreadsheet to see what my panels should be producing under full sun at noon, I can scroll the entire month and see what the highest PV Watts estimated numbers are and then compare my panels to those numbers when they're under full sun at noon. Again, not super accurate as I'm sure temperature and humidity in the air will make a difference..

    I'm curious as to the results others have when it comes to months or years of data.. Did you come in over the PV Watts estimations or under.... and by how much?
    While using PVWatts hourly output is technically a misapplication of the PVWatts model (Oh the humanity !!), the clear sky data from PVWatts hourly output will be anywhere from dead nuts to +/- MAYBE 10% or more different than what your array sees and does for any specific hour under corresponding clear sky conditions at your location and for any specific hour of the year. Still, what you, I and others do with such data in a way similar to what you describe is probably a good dart throw and maybe a bit better than that at estimating clear sky hourly system output.

    But, there are about a dozen or so variables, both within PVWatts that are user variable, some of which you'll need to root around and figure out what best represents the reality of your array, and external variables, some of which are weather related and others array related that PVWatts knows nothing about.

    Also keep in mind that, contrary to popular opinion and what you may think, about 90% or more (+/- a bit) of the irradiance values of the TMY values that PVWatts mostly uses are synthetic data, that is, most of the irradiance data is simulated/estimated/modeled. You'll need to dig into various NREL sources to find that factoid, and it doesn't make the data less useful or not representative of what's likely to happen for modeling of long term output, but it gospel truth and predictability are not among the claims made for either the data or the model, particularly for short(er) periods.

    See NREL doc. : NREL/TP-5000-63112, sec. 5 and 6 for more info.

    Another route you may take, but not for the feint of heart or technically less astute, NREL's SAM will give what may be a more representative but perhaps only a slightly closer dart throw on output. It uses what's mostly the same weather data, but it does allow more specificity with respect to system variables. Once dialed in, you can use it's output as a way to have PVWattts impersonate the SAM output by modifying PVWatts loss parameter and other PVWatts variables to match the SAM annual output. The differenc in hourly output will still remain, but probably be a bit less on an RMS difference. Doing so for my situation, for example, results in a PVWatts system loss parameter of about 6-8 % or so. Your results will vary.

    Comment


    • #3
      Sensij described this well in another thread I can't find. I'll try to summarize this then let him correct me. TMY is the weather model that PVWatts and SAM uses by default. A TMY year is built by taking 20 or 30 years of collected weather data. Holes in data collection [more likely from irradiance than temperature, etc. as JPM suggested] from each station are filled in with reasonable values. Then, it is analyzed by month. Say you have 30 Januarys to pick from. They will look at each January individually with various metrics and compare to the group as a whole. First the most non-typical month is discarded, then the remaining 29 Januarys are looked at and the most non-typical month is discarded. And so it goes. Finally the month that remains is 'the January' and is inserted in its entirety into the model. Once all of the months are selected, the first couple days and last couple days of the month are 'smoothed' a bit to match the month preceding/following.

      So you can see, the monthly outputs are probably the most reliable for TYPICAL, the weeklys less so, the dailys and hourlys almost meaningless for TYPICAL.

      That said, although they are not very useful for 'typical', the dailys and hourlys are still pretty good for finding best and worst. You can copy the hourly output column to a blank column, then sort it. Do you really want to size something just to catch the top 20 hours in the model? Perhaps not, count down 20 rows and see what the value is. If that results in a different perspective, then design to it and know that you have clipped 20 of the 8779 (or whatever it is) hours in your year.

      Sensij, - help me out here. Am I close enough? Not to exclude all of the rest of you that also know this stuff...

      One interesting tidbit - a couple of calendar years in the 90s were completely disqualified due to volcanic activity that had global implications long enough to scrap the entire year.
      Last edited by AzRoute66; 08-26-2017, 06:38 PM.

      Comment


      • #4
        I just found PV Watts a few days ago due to this forum. From what I have seen of my production and what PV Watts says for my install, it is going to be just about right for the year - maybe a few hundred watts difference. IMS, it is close to what the installer said as well. Perhaps they used it for their estimates.

        Comment


        • #5
          Originally posted by DrLumen View Post
          I just found PV Watts a few days ago due to this forum. From what I have seen of my production and what PV Watts says for my install, it is going to be just about right for the year - maybe a few hundred watts difference. IMS, it is close to what the installer said as well. Perhaps they used it for their estimates.
          I'd bet money on it...

          Comment


          • #6
            UPDATE:

            We "officially" activated the array on Aug 22nd and started keeping records on the 23rd. I had downloaded the PV Watts hourly spread sheet to glean information from it for various purposes.

            I modified the PV Watts spreadsheet and had excel add up the hourly figures for each day's predicted production total, then I look at my SMA inverter interface to see what was actually produced. I also modifed the PV Watts to "average" what is expected against what we actually got for the time frames.

            After 10 days, PV Watts is so close to our actual production is freaking scary.. PV Watts says we should be averaging 31.59 kW per day, our actual is 31.67 kW.. (100.24%)... Dang that is close.
            The actual figures for the individual days vs. what we actually get for that day, almost never agree obviously... but summing and averaging the data for the 10 day period seems to be spot on what PV watts calculated.

            Our house uses between 20 and 22 kWh per day.. we seem to be banking about 10 kWh per day on average. We did 41960 watts today! Our highest has been 44640 and our lowest was a miserable 13690...
            Last edited by Murby; 09-02-2017, 12:30 AM.

            Comment


            • #7
              I just looked back at my first two months of production vs my PVwatts estimates. Using the production numbers from PVOutput (since everyone agrees that the numbers SolarEdge gives are high).
              Month PVWatts PVOuput Difference
              July 1074 1134 +60 (5.6%)
              August 1013 1096 +83 (8.2%)

              So I'm overproducing by a fair amount. I have a split array and have modeled it as such in PVWatts (two separate estimates which were combined)

              Comment


              • #8
                Originally posted by jasonvr View Post
                I just looked back at my first two months of production vs my PVwatts estimates. Using the production numbers from PVOutput (since everyone agrees that the numbers SolarEdge gives are high).
                Month PVWatts PVOuput Difference
                July 1074 1134 +60 (5.6%)
                August 1013 1096 +83 (8.2%)

                So I'm overproducing by a fair amount. I have a split array and have modeled it as such in PVWatts (two separate estimates which were combined)
                If your output matches PVWatts it will be more coincidental than anything you or your array are doing. it'll probably be close, but thinking any more than that will be like thinking the weather that PVWatts uses is identical to the weather you had during the periods you estimated measured your output. Besides, most of the TMY stuff is itself modeled data. Probably good enough for its purpose, but again, mostly modeled and not actual measured data.

                Comment


                • #9
                  Originally posted by Murby View Post
                  UPDATE:

                  We "officially" activated the array on Aug 22nd and started keeping records on the 23rd. I had downloaded the PV Watts hourly spread sheet to glean information from it for various purposes.

                  I modified the PV Watts spreadsheet and had excel add up the hourly figures for each day's predicted production total, then I look at my SMA inverter interface to see what was actually produced. I also modifed the PV Watts to "average" what is expected against what we actually got for the time frames.

                  After 10 days, PV Watts is so close to our actual production is freaking scary.. PV Watts says we should be averaging 31.59 kW per day, our actual is 31.67 kW.. (100.24%)... Dang that is close.
                  The actual figures for the individual days vs. what we actually get for that day, almost never agree obviously... but summing and averaging the data for the 10 day period seems to be spot on what PV watts calculated.

                  Our house uses between 20 and 22 kWh per day.. we seem to be banking about 10 kWh per day on average. We did 41960 watts today! Our highest has been 44640 and our lowest was a miserable 13690...
                  JPM said it perfectly. I would only add dont set yourself up for disappointment or worse, worry about system malfunction, if the numbers start to deviate by very very large amounts per day; even per month or longer like several months in a row pvwatts deviation happens when weather is suspected.

                  That is great you are banking so many kwhs. It is 109 here right now, AC running all day, not much excess for us and the heat has us down to 30kwh daily with 7.28kw south facing minimal shade. Sorry if so many of us lost that initial excitment a while back. Welcome to the solar community.

                  Comment


                  • #10
                    So far after17 months, I have produuced just a little over the PVWatts estimate all but one month (Jan). Probably something like 2% or so on average. With that said, I did leave the system losses at the default level which I think is a little high.

                    Greg

                    Comment


                    • #11
                      Originally posted by Murby View Post
                      When I designed my solar system, I used NREL's PV Watts calculator to determine sizing. My solar system is on its 5th full day of energy production and I'm watching what was calculated (via PV Watts) vs what I'm really getting..
                      Obviously, five days of data is not even remotely close enough to make any determinations but there are certain things I can decipher. For instance, when I scan the PV Watts hourly data spreadsheet to see what my panels should be producing under full sun at noon, I can scroll the entire month and see what the highest PV Watts estimated numbers are and then compare my panels to those numbers when they're under full sun at noon. Again, not super accurate as I'm sure temperature and humidity in the air will make a difference..

                      I'm curious as to the results others have when it comes to months or years of data.. Did you come in over the PV Watts estimations or under.... and by how much?
                      Further to my 09/02/17 post:

                      I keep pretty good track of my system's output, whatever it may be. One of the several numbers I have is a running total of my system's actual 31 day output vs. what PVWatts "thinks" the output of my system "ought" to be for the same 31 day period. I juggled the PVWatts system loss parameter such that the PVWatts output matches both my own model and SAM's output so that all 3 are within 1% or so of one another in terms of annual output. That forces a PVWatts system loss parameter of ~ 8%. I have late afternoon shading that decreases daily system output by very approx. 3%-5%, depending on time of year.

                      Using daily output data from 10/18/2013 through 08/10/2017, I have 1,380 data, each one being the ratio of 31 days of total actual system output to the same 31 calendar days of PVWatts estimated total system output, with that PVWatts output being massaged so that it matches SAM's output, and my model's estimate as described.

                      The distribution of the data is as follows: The mean ratio of actual system output to PVWatts' modeled output for 1,380 ea., 31 day periods is 0.961. That is, on average, my system generated 96.1 % as much as an adjusted PVWatts thought it would for any consecutive 31 day period. That seems reasonable to me, given the late afternoon shading that PVWatts knows nothing about, but I'm pretty confident of being ~ 3% to 5% as mentioned above.

                      BUT, and this is where I'm going, while the average ratio was 0.961, the minimum ratio for the same 1,380 running 31 day periods was 0.619, and the maximum ratio was 1.311. The population standard deviation for the 1,380 data was 0.098. In other words, The data seem to have somewhat of a normal or Gaussian distribution.

                      Put another way, If I assume 4% shading, the PVWatts long term average (if a scant ~ 3.8 years can be considered "long") looks pretty dead nuts to what my system actually did, but I had to make the PVWatts model behave closer to what SAM might estimate. Actually, given how much the "variables" can vary, I'm surprised it got that close.

                      However, the interesting part, at least to me, is that looking at any one 31 day period, that periods output could be off from the PVWatts model by, for my data, from between (1.311-.961) = 35% high, to (.619 -.961) = -34.2 % low.

                      FWIW, those extremes are not too far off from what the PVWatts literature and help screens suggest is the likely, or at least possible, variation when comparing actual output to PVWatts modeled output for monthly data, with most of that variation, they suggest, being due to variation in the amount of insolation received over any 30 or so day period.

                      PVWatts is a model for long term annual output. Using it for shorter periods is not within its capabilities. See the PVWatts help screens or the TMY manual for more information.
                      Last edited by J.P.M.; 09-04-2017, 01:36 PM.

                      Comment

                      Working...
                      X