Announcement

Collapse
No announcement yet.

Historical data analysis.

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

  • Historical data analysis.

    In the day job I work in enterprise big data. I was bored at the weekend and decided to point own of those tools at my solar panel data.

    Call it an academic test panel, 50W.

    What I have however is high resolution data for it's panel voltage, output current and output power for nearly a year.

    The first question I asked it. ("it" being my coding).

    "Is it sunnier in the morning or the afternoon?"

    The answer, should really have been, "neither. It's sunniest at noon you idiot.", but that was not the results I got. At first it seemed to and I made a sad face as it would be complete statistical insignificance. However my lips up turned and my mouth opened into an "Oohhhh....." when I realised that "solar noon" is not at GMT "noon" for me. I am 8* West. My solar noon is around 12:35. However my peak solar output was 11am and 12pm almost neck and neck as opposed to 12pm to 1pm.

    It means when I put the off grid panels on the garage roof I can bias them a few inches to the west to get the morning sun earlier from teh house shadow, rather than the evening sun over the hedge.

    Of course this could very well also be the impact of the battery running full charged for half of the summer when it only gets topped up in the morning sun and then the panel goes open circuit.

    Any other cool analysis I can do while I have this?

    I did for example, run my mains consumption figures through some analysis. I was looking for individual appliances by grouping "deltas" when power use increases or decreases and then trying to pair them up to actual loads. That ended up suffering either too much noise with low granularity or being able to say if it's a 100W load or a 500W load and about nothing else. Still, with slightly less lofty goals I hope to return to this and at least pick out one or two high power devices, like the oven or dishwasher and see if I can then exact their individual consumption month by month.

  • #2
    When I started making graphs, I needed to plot solar noon (from a
    chart), that varied as much as 20 minutes from standard time. Then
    the graph would be symetric. Bruce Roe
    Last edited by bcroe; 03-31-2023, 10:35 AM.

    Comment


    • #3
      The more depressing figure was much easier to get.

      Solar variance.

      Average panel output / Panel rating

      My previous "go to", "finger in the air" was 10:1 for a 24 hour load. 100W 24 hour load = 1kW panels.

      Turns out that my solar variance is about 18-20:1.

      Comment


      • #4
        Under the clouds here average inverter output
        approximately equals panel rating/28
        Bruce Roe

        Comment


        • #5
          Near the Atlantic in Canada and Northern US, the solar afternoon is usually sunnier than the solar morning. Fog occurs mainly in the morning, and dissipates by noon.

          Comment


          • #6
            It's this kind of stuff that makes it non irrelevant. Removing the double negative, it's perfectly relevant to assess hourly solar availability, seasonal or otherwise.

            (Local micro-climate for a particular roof can make a sizable difference).

            I am on a campaign to get far, far more detailed than that. I am on the scent of solar power estimates from weather models, which correctly ingested can be used to assess the average impact of local micro-climates and include impacts such as seasonal or monthly cloud opacity.

            Comment


            • #7
              Originally posted by venquessa View Post
              It's this kind of stuff that makes it non irrelevant. Removing the double negative, it's perfectly relevant to assess hourly solar availability, seasonal or otherwise.

              (Local micro-climate for a particular roof can make a sizable difference).

              I am on a campaign to get far, far more detailed than that. I am on the scent of solar power estimates from weather models, which correctly ingested can be used to assess the average impact of local micro-climates and include impacts such as seasonal or monthly cloud opacity.
              How involved do you want to get into the technical/engineering aspects of solar energy and its applications ?
              Meant as a respectful observation; It reads to me like a lot of what you're describing is available for the looking and some reading and you may be reinventing the wheel. Or, maybe I'm not understanding what you're writing.
              One example only; Solar availability is so relevant it has been around as a science for > 100 years and about beaten to death for the last 50 years and sliced and diced more ever since alternate energy took off.
              Another: Solar geometry calcs and algorithms are commonly available to calculate solar position with very high degrees of accuracy. It's very easy to calculate solar noon, or any solar time to within a fraction of a second and the solar position relative to any spot on the earth to within 0.003% or so of 1 degree if a way to calculate local atmospheric parallax is available.

              Comment


              • #8
                Originally posted by J.P.M. View Post
                Another: Solar geometry calcs and algorithms are commonly available to calculate solar position with very high degrees of accuracy. It's very easy to calculate solar noon, or any solar time to within a fraction of a second and the solar position relative to any spot on the earth to within 0.003% or so of 1 degree if a way to calculate local atmospheric parallax is available.
                One factor I find lacking in the models is wind. Near the Atlantic around 45 degrees, spring and fall are much windier than summer. I've noticed SAM, which is the model used by PVWatts and OpenSolar, underestimates spring production, and overestimates summer production. SMAs simulator, Sunny Designer, shows a similar bias. AFAIK TMY weather files include wind, and the temperature of PV panels based on insolation, tilt, and wind is well studied. Between wind and snow coverage, I've had to look at over a dozen other local systems and read a few research papers before I figured out how system output typically varies from the modeled estimates.

                Comment


                • #9
                  Originally posted by venquessa View Post
                  It's this kind of stuff that makes it non irrelevant. Removing the double negative, it's perfectly relevant to assess hourly solar availability, seasonal or otherwise.

                  (Local micro-climate for a particular roof can make a sizable difference).

                  I am on a campaign to get far, far more detailed than that. I am on the scent of solar power estimates from weather models, which correctly ingested can be used to assess the average impact of local micro-climates and include impacts such as seasonal or monthly cloud opacity.
                  As far as I know the climate is changing which will produce different cloud covers that historical weather data may not be able to provide the correct variables for you to predict the future

                  Also it takes only one event (storms, volcanic emissions, etc) that can really skew the amount of future sunlight for a lot of areas.

                  Look at the amount of snowfall on the West coast this year.
                  Last edited by SunEagle; 04-28-2023, 08:21 PM.

                  Comment


                  • #10
                    Originally posted by nerdralph View Post

                    One factor I find lacking in the models is the effects of wind. Near the Atlantic around 45 degrees, spring and fall are much windier than summer. I've noticed SAM, which is the model used by PVWatts and OpenSolar, underestimates spring production, and overestimates summer production. SMAs simulator, Sunny Designer, shows a similar bias. AFAIK TMY weather files include wind, and the temperature of PV panels based on insolation, tilt, and wind is well studied. Between wind and snow coverage, I've had to look at over a dozen other local systems and read a few research papers before I figured out how system output typically varies from the modeled estimates.
                    About the most variable and least reliable parameter in solar design and/or modeling is the wind. Most all models in any engineering discipline suffer from a lack of more precise ways to account for wind vectors and their effects.

                    Among other things, once upon a time I did a fair amount of structural design, and stress and failure analysis for power and process equipment of all sizes and types that went into power plants, refineries and chemical plants. Wind design is a subset of that. It's a fascinating subject but something of one of the black arts with respect to structural design, especially when dealing with vibration analysis and particularly of tall towers with high slenderness ratios, or flow induced vibration analysis in tubular heat exchanger products like the ones that failed at San Onofre N.G.S which was a mess that should never have happened - but that's off topic.

                    I'm more familiar with thermal design and heat transfer including natural (gravity induced) and what's actually treated like forced convective heat transfer with the force coming from wind that unlike more civil and predictable flow regimes (like turbulent flow in a conduit or channel for example) that have more/less steady flow, wind is unreliable and chaotic which is one reason it's so hard to get a handle on. That's the type encountered in solar and other flat plate designs with respect to wind.
                    Unless it's a standard situation done without variation like in a process heat exchanger, any model, algorithm or empirical correlation that gets within 10% of reality half the time it's considered a success by heat transfer slugs. Honest. dirty little secret.

                    I've done dimensional analysis on thermal systems for a living and from that it's been a busman's holiday coming up with empirical correlations for my system based on a boatload of analysis and measurements. After about 10 years of that I satisfied myself that I've got my system dialed in pretty good. Basically, I wound up confirming most all of the heat transfer stuff I ever learned that's applicable for this application.

                    I'm not at home just now but if you want to, holler back and I'll post my wind correlation coefficient and point you to some similar info I get back to San Diego. You may have seen some or all of the published stuff.
                    Just know that to get something that makes sense out of it for an application, you'll need a way to get GHI at your array and a way to convert that to P..O.A. irradiance as well as a way to get a reasonable wind vector at the array.
                    I've had a Davis Pro II+ weather station located about + about 4 ft. north with the anemometer about 6 " above the north edge of the array.
                    BTW, ever notice where the anemometers on instrumented panels and arrays are usually located ? The ones I've seen and seen photos of are usually far removed from the panel or array. By far away I mean more than a couple of meters away and more than a meter or so above or below the array.
                    Wind effects are VERY local. Ever walk around the corner of a building and get a blast of air ? IMO, that type of variability and the need to be very close to what you're trying to measure is one of the other several, big and fixable reasons why the wind vector and coefficient is as unreliable as it seems to be for all types of equipment. Measuring wind is more than a wet finger stuck somewhere.

                    With respect to SAM's validity as a model, I haven't found the seasonal accuracy variation you noted. I don't doubt it. I just haven't seen it in my application.
                    I've heard it suggested that there is regional variation in the model's validity using the old TMY2 and 3 data sets, which, given that most all of that weather data is modeled from the 26 old SOLMET sites all over the U.S., I'm surprised it gets as close as it does. (See the TMY manual for more details).
                    I don't know and haven't heard how or if the SolarAnywhere data will give more or less or the same model stability, either regional or temporally than the TMY data. As best as I can guess for my application, Sam's seems about, maybe, +/- a couple % or so of on production over 365 days after accounting for normal and daily weather and irradiance variation. I have not noticed a seasonal variation. I have noticed that SAM seems to underperform on system output a bit on some hourly values in mid-morning and midafternoon if/when my array is more heavily fouled. Long tale of how I came to that conclusion. A lot of repetition, statistics and regression analysis.

                    Comment


                    • #11
                      The wind has been of more interest to me this past month as I never installed the solar panel yet. It's sitting, literally sitting on the ground leaned against the garage wall. It still produces more than it's rating on occasion. The garden is very sheltered which means the wind doesn't get down into it much. So 50km/h winds haven't moved it yet. The risk is the gusts and in higher winds the air in the garden can become very unstable wish little gustlets blowing chairs over and stuff. Thankfully the rails/hooks arrive this week and my spark is going me a hand to mount it on the garage roof.

                      At the moment all I was hoping to calculate was my own solar variance. I have data but it doesn't go back a year yet and to make it more difficult the panel size changed in the middle of it.

                      For April, for example.the panel averaged 19.1W, as it's a 330W panel that works out around 17.2 W/W. If I want a 100W 24/7 load, I need 1720W or more panels. In April this year anyway.

                      I was hoping to find some freely available data for the same going back a while.

                      Comment


                      • #12
                        Originally posted by venquessa View Post
                        The wind has been of more interest to me this past month as I never installed the solar panel yet. It's sitting, literally sitting on the ground leaned against the garage wall. It still produces more than it's rating on occasion. The garden is very sheltered which means the wind doesn't get down into it much. So 50km/h winds haven't moved it yet. The risk is the gusts and in higher winds the air in the garden can become very unstable wish little gustlets blowing chairs over and stuff. Thankfully the rails/hooks arrive this week and my spark is going me a hand to mount it on the garage roof.

                        At the moment all I was hoping to calculate was my own solar variance. I have data but it doesn't go back a year yet and to make it more difficult the panel size changed in the middle of it.

                        For April, for example.the panel averaged 19.1W, as it's a 330W panel that works out around 17.2 W/W. If I want a 100W 24/7 load, I need 1720W or more panels. In April this year anyway.

                        I was hoping to find some freely available data for the same going back a while.
                        I believe you would benefit from an understanding of the basics of solar energy.

                        Try this website : pveducation.org . Lots of background basic stuff about how PV works.

                        Or, for about the best text on the solar energy resource : "Solar Engineering of Thermal Processes" by Duffie and Beckman. Aside from the title, it has about the most concise treatment of the basics of how PV works I've seen. Prior editions are online as a PDF.
                        It also contains extensive treatment of wind effects on any flat plate or a PV panel or a thermal solar collector's solar collector's performance.

                        For more background on solar energy and what's being done presented in what's maybe a bit more basic try: www.nrel.gov. You'll also find something called PVWatts there which besides being good tool for the preliminary design and sizing of PV systems, has a lot of info on the basics contained in the info and all the notes.

                        All the above is free.

                        If you're looking for information that goes back awhile, check out "A Golden Thread, 2500 Years of Solar Architecture" by Butti and Purlin, That's not free but it is a lot of interesting history.

                        Good Luck.

                        Comment

                        Working...
                        X