A preliminary reservoir balance calculation for El Hierro.

There are a number of key variables related to the Gorona del Viento (GdV) project that we have no data for, basically because GdV has declined to send us any. One of the most important is the reservoir balance, which up until recently we have only been able to guess at. However, thanks to the valued participation of Rainer Strassburger, our man on El Hierro, we now have some recent dated photographs of water levels in the upper and lower reservoirs. Are they good enough to allow us to make meaningful reservoir balance estimates? This post looks briefly into this question and concludes that they are.

GdV upper reservoir at the pre-construction stage, image credit vistaire

First a summary of results. Note that all numbers are approximate.

1. Between February 20 and March 28, 2016, there was a net transfer of 20,000 cubic meters from the upper reservoir (UR) to the lower reservoir (LR), with 40,000 cubic meters transferred from the UR to the LR between February 20 and March 17 and 20,000 from the LR to the UR after March 17.

2. There is an encouragingly close balance between the transfer volumes, indicating that photos of water levels can be used to estimate reservoir volumes to within acceptable limits of error.

3. The two reservoirs between them presently contain about 90,000 cubic meters of water, or less than 20% of GdV’s total combined reservoir capacity of 535,000 cubic meters. As of March 28 both reservoirs were only 15-20% full.

And second a brief description of the procedures I used to obtain these results.

One of the notable things about the GdV reservoirs is that they contain narrow terraces of white-painted concrete (?) blocks that divide them into three segments:

Figure 1: Upper reservoir under construction, showing white concrete block terraces (click to enlarge)

The primary purpose of these blocks is probably to anchor the liner, but the spacing is intriguing. It’s almost as if they divide the reservoir into three layers of equal volume, with the top line of blocks representing 100% full, the second 67% full, the third 33% full and the lowest empty. Detailed geometric analysis of reservoir shapes would be needed to confirm this, but I didn’t have the time to do any so I accepted the 33/33/33 argument as a working hypothesis and went from there.

The next step was to measure water levels relative to the white lines. I did this by copying Rainer’s images into Microsoft “Paint”, identifying common measurement points on all of them (yellow line) and recording Y-axis pixel values. Here’s the latest result from the upper reservoir :

Figure 2: Y-axis pixel counts, upper reservoir, March 28 2016, 157 hours

The water volumes that correspond with the first three pixel counts are 385,000, 257,000 and 128,000 cubic meters according to my 33/33/33 assumption, so next I plot these volumes against the pixel counts to see whether I get a straight line, which if my 33/33/33 assumption is correct I should:

Figure 3. Pixel counts versus assumed 33/33/33 reservoir volumes, upper reservoir, March 28 2016, 1537 hours

The line shows a slight dogleg but isn’t bent enough to cause me to discard my 33/33/33 assumption. So I proceed.

The next step is to estimate what volume the 264 pixel count at the water level represents.  I did this by adjusting the volume until I judged that the point fitted the projection of the Figure 3 plot as closely as possible. In this case 60,000 cubic meters (black) gives me a slightly better-looking projection than 50,000 cubic meters (blue) or 70,000 cubic meters (red), so I accept the 60,000 cubic meters number while recognizing that I’m probably looking at a +/-10,000 cubic meter estimation error.

Figure 4. Pixel count at water level fitted to 33/33/33 reservoir volume plot to estimate water volume, upper reservoir, March 28 2016, 1537 hours

And here’s the reservoir balance for the six days for which photos are available after all the volume measurements were to hand:

Figure 5:  Volumes of water contained in lower, upper and both reservoirs combined, February 20 to March 28, 2016. The open red circle in the lower reservoir plot is interpolated. No photo is available for March 14.

I was pleasantly surprised by these results. The LR fills and the UR empties, the UR fills and the LR empties, the volumes are comparable and the combined stored volume stays the same within 10,000 cubic meters, or 2% of the combined reservoir volumes. With a more sophisticated estimation approach than the one I used it should be possible to come up with some fairly exact numbers.

This entry was posted in Energy and tagged , , , . Bookmark the permalink.

129 Responses to A preliminary reservoir balance calculation for El Hierro.

  1. singletonengineer says:

    Why are we, external observers, forced to deduce this data? Why is GdV so reluctant to provide real time data to demonstrate the technical, commercial and environmental success of their publicly funded project?

    The short answer must be that the project is an engineering and commercial flop. Otherwise, they would have been broadcasting their detailed results far and wide.

    I have looked for and I have not yet found any single project on Earth that demonstrates, unequivocably, that anywhere on our globe, there is a project that demonstrates the value of solar photovalatic, solar thermal or wind unless supported by a combination of cash subsidies and of conventional electricity generation technologies such as coal, natural gas and nuclear. Come to think of it, how “natural” is natural gas when it is derived through fracking, processed on land, compressed, transported and ultimately delivered to gullible populations? Whenwill the industry agree to cooperate with those who seek to study the negatives as well as the positives of this somewhat unpopular energy source?

    El Hierro is only one of many that fail commercially and environmentally. Is there a single example of wind or solar anywhere that is not an example of failure?

    When will the fantasies about environmental benefit and about other attributes of unreliable weather-reliant so-called sustainable energy sources (are they, really?), be discovered by the population at large? Current news reports that Spain is withdrawing cash support from so-called “renewables” is sobering. Large companies are going bankrupt. When will those who place their trust in renewables, wind and solar, be those same folk who decide to adopt rational strategies that include all available energy sources?

    I say this, because the pro-renewables folk will have a lot to answer for when the day of reckoning comes; the day when coastal cities no longer function reliably due to rising sea levels. I truly feel for the poeple of Florida and Miami. And, of course, of London, Venice, New York and a thousand other cities.. They will be among the many cities, large and small, that will soon realise the magnitude of the problem and which will soon seek special support for their flooded community services and transport systems.

    If recent published predictions of 1 metre of sea level rise by 2100 are true, will finding somebody to blame for the last 40 years’ delay and inactivty help? Of course it won’t. We will need to find those who are prepared to work to avoid the problem, much more than we will need to be able to identify those who failed in their civic duty and to do their level best to do something about it.

    Where is the community service organisation I am a member of in this debate? I’m a Rotary club member. Will Rotary International take a position in the global energy and stand its ground? I doubt it.


    • The short answer must be that the project is an engineering and commercial flop. Otherwise, they would have been broadcasting their detailed results far and wide.

      The good results have been broadcast but the less good ones haven’t.

      What seems to have happened is that the project went ahead in a fit of green euphoria before anyone had drawn up any engineering plans or even performed the most basic checks to see whether the hydro-wind system was going to work. Maybe more about this next week.

    • Tom Bates says:

      The current ocean rise trend per NOAA tidal gauges on land not moving up and down is 3 inches in 100 years, Johnston Atoll and Sydney are examples. Miami has a land subsidence problem with an agency which is supposed to fix it but does not seem to be doing anything unlike Galveston Texas which has a similiar agency and problem. New York is not sinking into the sea but has a large portion of the city built at essentially sea level. A hurricane in the early 1800’s as large as the one which hit the area recently caused no damage because people had not built on sand dunes unlike later on something which is affecting a number of east coast states. You do know that antarctica and Greenland are gaining land ice not losing it so the ocean rise is unlikely to increase to the one meter number. CO2 forcing has actually been measured from 2000 to 2010, it was .08F increase with a 9 percent CO2 gain, 2/10th’s of a watt. If the CO2 doubles to 18 percent gain from 2000 the increase would be .0848F from 2000. It is unlikely CO2 is the cause for present warming considering NASA predicts 25000 years of warming from changes in earths tilt and orbit. The numbers they are using to push their agenda are surface warming numbers from Giss and Hadcrut. Hadcrut is mostly Giss. Giss is 66 percent an estimate so less than half the numbers are an actual measurement. RSS and STAR upper atmosphere measurement show no warming. Finally you should consider the medieval warming period when it was warmer that today and the depths of the little ice age when it was colder than today, none of that change was caused by CO2 which leaves changes in solar gain from changes in solar output.

    • Rainer says:

      “I have looked for and I have not yet found any single project on Earth that demonstrates, unequivocably, that anywhere on our globe, there is a project that demonstrates the value of solar photovalatic, solar thermal or wind unless supported by a combination of cash subsidies and of conventional electricity generation technologies such as coal, natural gas and nuclear”
      Just look at the places without grid access and you will find more you can imagine.
      El Hierro only:
      tracking of volcanic activities
      seismeter, GPS stations…..

  2. Since this post was written Rainer has posted April 1 photos of UR and LR water levels which according to my crude algorithm show ~10,000 cu m added to the UR and ~5,000 cu m removed from the LR since March 28 – another close balance.

    Unfortunately, however, Rainer will be going back to Germany in a few weeks and we will lose his invaluable photographic services. Energy Matters is therefore soliciting applications for the position of unpaid chief reservoir water level photographer to carry on the good work through the summer months. 😉

  3. Flocard says:

    Good work Roger That is what we need at this point. This leads me to the following questions :
    When one considers the total water level (black line), it is OK to say that

    1) if the curve moves up by about 15000m3 as betwwen March 17 and March 24, it means either that over this specific week it has rained a lot with little evaporation or a large amount of water has been bought from the El-Hierro water council. Within the second hypothesis this corresponds to 2000m3/day. This is a significant fraction of the capacity of the Cangrejo desalination plant working steadily at the service of GdV.

    2) if it is going down by about 18000m3 as between March 10 and March 17, it means either that there was a lot of evaporation or a loss in the ground or duming fresh water to sea or that a volume of about 2500 m3/day was sent to irrigation.

    It is hard imagining that brand new reservoirs are already leaking. In addition if they would do they would do it at a constant rate which would certainly generates a steady down goung trend which does not show up. Thus I do not retain this hypothesis.

    According to the El-Hierro water council over a year the total evaporation from both reservoirs is 149000m3 thus an average of less than 410 m3/day (probably less than that in March). I keep 400 m3/day as an average value for March

    It is difficult for me to imagine that GdV would be voluntarily or inadvertently dumping fresh water into the sea. (on the other hand the logic of GdV sometimes escapes me)

    If the black curve is correct a contribution from the upper reservoir to the irrigation system at a rate exceeding 1400 m3/day should have been observed bettwen March 10 and March 17. Can rainer comment on that ?

    For the period March 17 to March 23, was it raining a lot over the island enough to justify the replenishing of the ensemble of the two reservoirs? If not, keeping in mind an evaporation rate of about 400m3.day can’t be we investigate if water at a rate of 2000+400=2400m3/day has been bought by GdV ?


    • Hubert:

      Thanks for your detailed and reasoned response. However, my interpretation of these results is quite different – revolutionary even 😉

      According to the REE “hydro” values approximately 500,000 cu m of water was pumped up the hill between February 20 and March 28 (I’d appreciate it if you could check my numbers on this). Since this water had nowhere to go when it got to the top of the hill it must have come back down again, in line with our previous conclusions. This gives us a total of around one million cu m of water circulating through the hydro system, a volume that exceeds reservoir transfers by a factor of maybe twenty.

      Which leads me to suspect that the reservoir transfers weren’t planned. They just happened. They are to all intents and purposes “noise”.

      And why should GdV not care how much water gets transferred between the reservoirs?

      Because so long as it doesn’t rupture the liner the amount of water in the reservoirs has become irrelevant. All the evidence suggests that GdV has quietly given up on the idea of using them for energy storage and is now using the hydro system purely as a dynamic resistor to shed surplus wind power, match generation to demand and generate a little hydropower from time to time. And provided the penstocks are full it can do this by circulating water up and down the hill even when the reservoirs are empty. The concept is the same as your car’s cooling system except that it sheds energy rather than heat, and all you need to do to keep your car’s cooling system running is to top up the radiator from time to time. You don’t need a big coolant tank on the roof.

      And if this interpretation is correct then any resemblance between the GdV project as originally conceived and as it now functions has indeed become entirely coincidental.

      • Rainer says:

        Turismo el hierro????
        The first Island on the world with a 600m kids waterworld……

      • Flocard says:

        I am going to check the number you mention. I have just finished collecting REE data for the month of March. Give me one more day.

        In the mean time, I am ready to accept your interpretation that a very large amount of water (exceeding the combined capacity of the two reservoirs) is going up and down not producing useful electric energy for the island. and producing may be stability to the grid.

        I just would like to point out that water circulating from one reservoir to the other and back can’t change the total amount of water once it is summed over both reservoirs which is precisely what your black line tells us.

        I was only discussing the black line. For me this line is only affected by either injection of water from the outside world (outside the hydraulic comonent of GdV) or ejection of water toward this outside world.

        Unless I miss something it can’t be affected by the kind of management you describe (which I believe to essentially correspond to reality)

        Note that I am only talking of much smaller volumes, typically 15 000m3. They could also correspond to uncertainties in your evaluation of the amount of water present in each reservoir.


        • the black line …. is only affected by either injection of water from the outside world (outside the hydraulic comonent of GdV) or ejection of water toward this outside world. Unless I miss something it can’t be affected by the kind of management you describe ….

          Hubert, unerringly you have gone straight to the weak point of my argument. If all GdV does is recirculate water through the penstocks there is no reason the levels in the reservoirs should change, but change they do. You see it clearly in the photos, and my pixel counts should be accurate enough to identify volume changes of less than 5,000 cu m. So what’s happening?

          The answer is – I don’t know. But one thing I do notice is that while water volumes in the UR and LR are strongly negatively correlated (R2 = 0.86) they are uncorrelated with the water volumes pumped uphill estimated from the REE numbers (R2 = 0.01 for the LR and 0.04 for the UR).

  4. Kees van der Pool says:

    “All the evidence suggests that GdV has quietly given up on the idea of using them for energy storage and is now using the hydro system purely as a dynamic resistor to shed surplus wind power, match generation to demand and generate a little hydropower from time to time”


    • Rainer says:

      That is it.
      Next visit i will look for white flags of technical knowledge surrender on the management house……

      • Kees van der Pool says:

        You better do that from a distance – I don’t think you are very popular at GdV right now. . . . . . . .

  5. Flocard says:


    I have no answer either concerning this connexion of the hydraulic component to the “outside word” whether by pipes (in or out) or by natural causes (leaks, evaporation, rain).

    Among the positive contributions to the black line, there are only two I can think of : rain or injection of water bought from the El-Hierro water council (I use the verb “buy” because in the contract signed with the Spanish state, there is the explicit mention that water for the initial filling of the reservoir had a cost for the GdV consortium. Thus I assume that later injections of desalinated water are also to be counted on the negative side of the GdV financial accounting)

    We could check on the effect of rain (by getting the information on the rainfall over a given period from the weather buro) and multiplying by the oustide surface of the reservoirs as given for instance by Googlemap

    Then by difference with your figures we would know the MINIMAL amount of water bought by GdV over a given period. I use the word “minimal” because at the same time we see the total (both reservoirs) amount of water growing, there is some evaporation (typically 400 m3/day) and possibly some water is being diverted for irrigation (I eliminate leaks into the ground and water dumped into the sea as too unrealistic even for a GdV type of management).

    From here now I’ll focuse on something that might (or might not) cause inaccuracies in your analayses although I do not know at this point how I could do better than what you have been doing.

    But even if we are not able to implement “improvements” we might be able to see in which direction they could affect the figures of your analysis.

    First I note that the rather flat black line is not generated by a measurement but is the result of two separate (independant) measurements of functions (red and blue) which are much more rapidly varying. The absolute uncertainties add up. The uncertainty on the black curve is larger than that on either the red of blue curve.

    To estimate variations of volumes based on pixel counting you use a linear approximation. It is true that locally any C1 function (C1 at least of course) may be approximated locally by a straight line. On the other hand as I said variations even if the final variations of the black line are not large those of the red and blue curves are not always small (local). It is uncertainty on these rapidly varying functions which matter.

    Could the linear approximation generate uncertainties sufficiently large to be visible at the level of your estimates ? Indeed the natural relation between height and volume in the kind of reservoirs we are are talking about here, is in general more quadratic than linear.

    First I would like to point out the document “Modelado de las caracteristicas de embalses y centrales hidroelectricas” by Juan Palomares Carralero, an Iberdrola engineer and in particular the figure 1 of his publication (I got the document from the web).

    To come back to El-Hierro, the pictures of the completed and empty reservoirs you have published, seem to show that their bottom is (almost) flat and their slopes are constant. This means that the relation between height of water “h” and surface of water “S” is linear (surface is growing regularly as a function of height) could be of the type
    Because the surface is the derivative of volume of water “V” with respect to height :
    we have the typical relation :
    V(h)=S0*h + 0.5*a*h^2
    (we know that V=0 when h=0).
    The documents for the lower reservoir tell us that hmax=13m. We can get S0 and S(13) from Googlemap. Thus we could deduce the value of “a” at least for this reservoir. (I have not found hmax for the upper reservoir.)

    The question that comes later is of course how can one relate relate your pixel counting with height of water “h”. Is there a linear relation between pixel variation and height of water variation ?

    To conclude on an optimistic note : suppose that after thinking it over and over, you decide that after all, the (anti)correlation between red and blue curve is not 0.86 but is perfect (1) and that the black line is flat over this month of data you have been anlyzing,
    then the GOOD NEWS is that
    we can numerically sort out the amount of energy sent to pumping (right now we only know the “minimal” amount of energy sent to pumping) and thus the amount of energy generated by the turbines (we only know the amount of energy which has been effectively used by the islanders, not the amount of energy recirculated (= wasted) inside the hydraulic component of GdV).


    • Hubert:


      According to the EH airport records 1.6 units of rain was recorded between Feb 20 and March 28, 2016. It doesn’t say what the units are but even if they are inches we aren’t going to see any visible impact on the reservoir water levels.

      Water injection/extraction

      I think we can be confident there was no significant injection into or extraction from the UR. The water here seems to have been stagnant for some time, there’s no evidence of any water having come down the sluice and the flow gauge on the flexible black pipeline leading to Valverde shows no change over the period of observation (and is now broken. The pipeline has also been cut off above the water level).

      I think we can be confident there was no significant injection into the LR either except maybe to compensate for evaporation. If there had been injection in the +10,000 cu m range I think we would see it in the sum of the LR and UR volumes (the “black line”).

      But the really puzzling point …

      … is why we see a close correlation between water leaving/entering the LR and water entering/leaving the UR when there’s no good evidence that any water ever entered or left the UR. All of Rainer’s images show water entering the LR but none of them show any entering the UR. The UR sluice where I made the pixel counts in fact looks as if it’s never been used. All I can think of is that the water somehow comes in through the drain at the bottom of the UR, but I’ve no idea how this might happen.

      And the really important point …

      …. is whether the reservoir balance means anything any more. If we accept that GdV has given up on energy storage then it doesn’t matter how much water is in the UR and LR. The storage capacity is always zero, or as close to it as makes no difference. (As to why GdV would have done this, it could be because they’re concerned that the reservoir liners, in particular the UR liner, could rupture if they add any more water, but I have no way of confirming this.)

      • Kees van der Pool says:

        If there is a ‘crossover’ where we think it is, pumped-up water would appear at the drain of the UR, not at the sluice, the sluice being higher up. If the Peltons kick in and supply back pressure, this may happen, as the pumps cannot be turned off immediately to prevent ‘reverse water hammer’.

        • If there is a ‘crossover’ where we think it is, pumped-up water would appear at the drain of the UR, not at the sluice, the sluice being higher up.

          I think I just failed Hydraulics 101 🙁

          Kees, could your “reverse water hammer” effect account for up to 50,000 cu m of water being transferred from the LR to the UR in a month, which is what the reservoir balance data show?

          • Kees van der Pool says:

            No idea – it all depends on the control algorithms. But if you look at the vigorous switching between pumping and hydro lately, not unlikely. Yesterday (4/2) is a good example, a ‘sawtooth’ hydro pattern.
            There is no need anyway (and would be very difficult) to ‘balance’ the levels exactly at the crossover point by turning the pumps off immediately after the Peltons kick in. There is no penalty involved other than an increase in the upper reservoir. According to GdV, they work at the same time just fine and therefore overflow into the upper reservoir will happen.

          • Kees: Good stuff! Now all we need to explain is how the ~50,000 cu m of water that went up to the UR drains back down again to the LR. One plausible explanation is that the water is deliberately allowed to flow down, since the less water in the UR the less chance there is of a liner failure.

    • Rainer says:

      “(I have not found hmax for the upper reservoir.)”

      upper eservoir: 12m
      lower reservoir: 15 m

      “Final configuration

      The final configuration of the hydro-wind plant is as follows:
      — Upper reservoir: Maximum capacity of 379,634 m3, 12 metres of maximum water level with maximum elevation of 709.5 metres above sea level, 2 mm high-density polyethylene (HDPE) geomembrane for waterproofing.

      — Lower reservoir: Storage capacity of 150,000 m3, rockfill dam with maximum height of 23 metres, 15 metres of maximum water level with a maximum elevation of 56 metres above sea level, 2 mm HDPE geomembrane for waterproofing.”


  6. Rainer says:

    since 2016-04-06 06:20 GDV Wind + Hydro only

  7. Rainer says:

    That i wanted to see since i observe!
    Wind Only!
    Gratulation to GdV management!!!!!
    2016-04-06 11:40 0 0 0 6.5 0 0 0 0
    2016-04-06 11:50 0 0 0 6.4 0 0 0 0

    • Thanks for the heads-up Rainer,

      What’s different about this test is that it’s the first one they haven’t done on a Sunday.

      • Rainer says:

        Wind only also???
        That for shure i missed to get data from

        • Greg Kaan says:

          Did you manage to check the pumping station during that interval, Rainer?
          The single net figure for pumping and hydro makes it impossible to know if they weren’t offsetting the pumps with the peltons.

  8. Rainer says:

    Sorry, no,
    did think seriously about,
    but have really problems with tooth pain going up and down -air.pressure-. Have to clear this out first.

  9. Rainer says:

    2016-04-06 17:10 Diesel Back
    Today: 17h without diesel
    Since July 2015: 128h, under 2% of 8760h yearly duty…

  10. Rainer says:

    …or just end of shift…..

    • Kees van der Pool says:

      Maybe that, too.
      I don’t understand why they turned on the diesel(s), the evening peak could have easily been taken care of by wind, still at 11 m/s, and a bit of hydro. Nighttime use is only 3MW or so. There would have been very little risk involved.

      • Rainer says:

        for sure no good promotion for GDV and El Hierro…….

        • Greg Kaan says:

          It is possible that the grid stabilization without the diesels requires much more manual intervention (varying pumping and pelton settings) to match demand with generation. Staffing would then be a limiting factor in when the diesels could be switched off.

          After all, this level of wind generation penetration really hasn’t been done before so any automated systems for the combined hydro + wind turbines will not be mature as the diesel governors.This is the sort of issue that would be resolved over time with gradual increase of renewables only operation as experience and confidence grows.

          • Greg: I think King Island, Tasmania, although smaller than GdV, is achieving similar levels of wind penetration using the same kind of procedures (dynamic resistors, battery storage, flywheels and of course diesel backup). Last November they claimed 100% wind generation for 30-something hours straight.

          • Greg Kaan says:

            The difference may simply be greater experience (KIRIEP has been running longer and adapting with different components with lessons learnt).

            The difference in the balancing components may also play a part – the pumping and pelton response may not be as linear to control inputs as the diesel UPS, resistive bank and battery. But I think experience is the key – GdV should improve as time goes by

  11. Kees van der Pool says:

    Hi Greg,

    They have been ‘at it’ for a while too at El Hierro – this paper is a detailed description of the integration of the different components, the reason for the various kinds of control schemes and their response times. They refer to Dinorwig for validation of the GdV model:


    This paper describes the initializing routine of the Peltons and hooking them directly into the grid to supply inertial stability:


    A neat, somewhat less detailed article that also touches on stability is here:



    • Greg Kaan says:

      Thanks Kees.

      I have read some of those before but as all engineers will know, extending from a modelled simulation does not guarantee knowledge of the final system.

      In particular, the first paper models the system loads as purely resistive and that is clearly not the case with induction motors so, since the windfarms are supposed to perform the reactive balancing, their behaviour will then be affected which is likely to have ripple effects throughout the system.

      The second paper models the GdV system mainly for the case where the wind farm trips rather than the “normal” operation (there was a similar paper linked by Roger in an earlier article). Grid stability under “normal” conditions appears to be assumed to be non-concern.

      As always, the devil is in the details.

      BTW King Island has had wind turbines since 1998 and battery backup since 2003 and most of the current equipment since mid-2002. I would not discount the system management expertise they have built up in the course of all this time


      • Kees van der Pool says:

        I’m sure there is exchange of information, however, the missing part on King’s island is the pumped storage facility.

        You wrote: ‘In particular, the first paper models the system loads as purely resistive and that is clearly not the case with induction motors’.

        The first paper models the penstock column + turbines, not the complete electrical system. The electrical load was ‘simplified’, done to make the (already complicated) frequency response analysis of the hydro part a bit less complicated: @4. ‘this assumption is usual in the case of power plants connected to isolated systems’.

        • Greg Kaan says:

          The electrical load was ‘simplified’, done to make the (already complicated) frequency response analysis of the hydro part a bit less complicated: @4. ‘this assumption is usual in the case of power plants connected to isolated systems’.

          And I am saying that the load model may no longer be valid due to this simplification not being accurate with the combination of wind turbines and induction motors, making the frequency stability more difficult than modelled.

          Models can be completely invalid if the wrong elements are simplified or linearised – after all, this is the heart of the debate on GHG emissions.

          The other assumption is that the models are completely correct and the GdV operators are incompetent. This would be the least likely case, IMO.

          • Kees van der Pool says:

            I think we may have a bit of misunderstanding here.
            You wrote: ‘making the frequency stability more difficult than modeled’.
            Do you mean the frequency stability of the grid?

          • Greg Kaan says:

            That is correct, Kees. The frequency stability I mentioned was for the grid. It was in reference to section “5.1 Normal Operating Conditions” where they summarise the system frequency deviations for the 2 modelled tuning criteria.

          • Kees van der Pool says:

            Greg, the paper models the ‘hydro’ part of the system, not the electrical load of the generator, which is represented by a resistor. It optimizes the ‘governor’ parameters in finding a compromise between well damped oscillating behavior and an acceptable speed response.

            This is a worthwhile exercise as the behavior of the hydro turbine is very much influenced by the characteristics of the water column, the effects of water inertia, water compressibility and penstock wall elasticity. Water inertia effect causes changes in the turbine flow to lag behind changes in turbine gate opening or closing. Pipe wall elasticity causes travelling pressure waves (‘water hammer’).

            For a tiny island grid with a single generating point and without long transmission lines, I think the substitution of a reactive load with a resistor does not make any difference in the way the primary frequency is regulated to the set value. If the generators collapse due to an excessive reactive load as the controller adjusts the mechanical power input to the turbine , it would be poor system design and not the way of regulating the RPMs of the generators that is the cause (VAR management with capacitors or switched capacitors come to mind).


      • Rainer says:

        this paper even got the rpm of the pumps:

        a new attempt to the soundfiles???

  12. jacobress says:

    The main question is: to what extent is the GdV energy project linked to the island’s water supply system. The link is there, obviously: they use sweet water, not salt water. They depend on the desal plants to fill the reservoirs.
    The fact that the reservoirs are almost empty, all the time, could be because of one of two reasons: 1. lack of water, 2.some defect in the reservoirs themselves, could be they are not yet fully operational.

  13. Rainer says:

    A partly local blackout 12:20???

    2016-04-07 12:10 6.2
    2016-04-07 12:20 4
    2016-04-07 12:30 4.8

    Generation Diesel Wind Hydro
    2016-04-07 12:10 1.8 0 0 6.2 0 0 0 -1.7
    2016-04-07 12:20 1.1 0 0 4 0 0 0 -1.1
    2016-04-07 12:30 0.3 0 0 4.3 0 0 0 0.2

  14. Rainer says:

    2016-04-07 12:40 Diesel OFF

  15. Rainer says:

    2016-04-07 12:50 Diesel ON

  16. Kees van der Pool says:

    I’m pretty sure there is mutual feedback.

    The component missing on King’s Island is the hydro part. The first paper does not model the system load but the penstock frequency control. To make the hydro analysis a bit less complicated, the electrical load was assumed to be resistive and inertialess (4).

  17. Rainer says:

    2016-04-08 23:50 Diesel OFF

  18. Rainer says:

    2016-04-10 00:50 Diesel on

    • Roger Andrews says:

      Thanks for the updates Rainer.

      What’s been happening is that over the last week or so GdV has been enjoying a period of strong and sustained winds – possibly one of the strongest and most sustained since operations began – and has been performing tests to see how long the plant can keep going without diesel under these conditions, which according to projections should allow continuous 100% renewables generation once the plant is running correctly.

      The test period began on April 6. Over the ~110 hour period between then and now diesel generation was maintained at around 1.8MW, output from the wind park was curtailed at around 6.5MW and demand was balanced by switching wind output between the grid and pumping with a little hydro generation thrown in. Test periods when the diesels were turned off were:

      0630-1650 on April 6. 10 hours 20 minutes duration. Reason for ending test unknown.

      0000 April 9 to 0040 April 10. 24 hours 40 minuted duration. Reason for ending test unknown.

      This gives a total of 35 hours of diesel-free generation over a ~110 hour period, or roughly 30% with 100% renewables generation. I haven’t downloaded the data yet but eyeballing the REE charts gives me about 80% total renewables generation over the test period, although much of this was a result of limiting diesel generation to 1.8MW rather than shutting it down altogether.

      There was, however, another problem. The El Hierro grid crashed at 12.10 on April 7, reason again unknown. The diesels shut down and had to be ramped rapidly back up from zero to 5MW to cover demand, and eleven hours passed before REE felt confident enough to return to the 1.8MW diesel/6.5MW wind operating mode.

      Mixed results, I would say, and still a long way to go.

      Rainer, Kees et al. Please feel free to add comments, and apologies if I’ve re-invented the wheel.

      • Euan Mearns says:

        I guess one of the key questions is whether the current windy conditions are the norm and the calm conditions some kind of anomaly. With 12 GW installed wind, sitting atop a volcano in the middle of the Atlantic Ocean, perhaps they expected to get 50% load most of the time?

        I’ve not managed to follow all the discussion. But do we have a ratio for synchronous / asynchronous supply limit?

        • Here’s wind. Note the month divisions are approximate:

          • Kees van der Pool says:

            The windspeed at the windfarm must be quite a bit higher than at the airport. During this last diesel-less interval the maximum speed at the airport was about 10 m/s.
            At 10 m/s the Pwind max is 3.4MW (11.5*1000/3375) (the nominal is 2.3 MW @ 15 m/s per Enercon E-70 E4 2.300 unit).

          • Euan Mearns says:

            I have visited the Canaries many times, normally in October. There’s always a windy side and a calm side of whichever island you are on. While a 30 day mean masks the short term variability of wind speed, as a matter of interest, what wind speed is required for the 12 MW of turbines to provide 6 MW of electricity?

            There are a lot of variables here that we could do to summarise into a ready reckoner.

      • Kees van der Pool says:

        This sounds about right. The curtailing starts to make a lot of sense in view of the water volumes in the LR and UR – no need to increase the volume of water that is pumped around and less strain on the components.

        • Rainer says:

          Sorry Kees, can not follow.
          If you have a storage and can fill it for free, not doing it and do not use the storage in the moment you can substitute diesel for we it is just wasting diesel and public money.

          • Kees van der Pool says:

            No water to fill it and yes, the whole pumped storage installation right now is just mainly a variable load to regulate supply and demand with occasionally some net hydro generation thrown in. A waste? yes, until they find water to fill up the LR and start pumping it energetically uphill and not around.

  19. Kees van der Pool says:

    @ Euan, April 10, 2016 at 9:23 pm (‘reply’ button AWOL)

    To get 6MW out of the array, you need about 12 m/s unless my math is off.

  20. Kees and Rainer: Dug out of my files is this XY plot of airport wind speeds and GdV wind farm generation versus a plot of expected output from a “typical” wind turbine. It roughly covers July through November 2015. It suggests that wind speeds up on the hill and at the airport aren’t that much different.

    • Kees van der Pool says:

      The wind generation in the time interval from 0000 April 9 to 0040 April 10 looks remarkably flat (‘curtailed’). The wind in this time frame varied from 8 m/s to 13 m/s. According to Rainer’s Enercon graph, the output at 8 m/s is about 3 MW to 4.5 MW. I’m just curious why we did not see more of a modulation.

    • nukie says:

      The problem is we do not know which way they curtail, if they regulate the wind turbines to provide roughly 50% of the possible output, if they curtail to 6,5 MW or if they switch of some turbines. For me it seems they do it differently at different times. If they switch of turbines, or limit them to a percentage of the possible output, it would look ike as if wind is about the same.
      So too much remains unknown, unfortunately. Also the relation of wind speeds might change with the direction of wind.
      But from the design side – if you look for a good place for a airport, you would choose one not too exposed to strong wind, to avoid desasters. If you look for a wind farm site, you’d look for one most exposed to wind. So if airport and wind farm have the same wind in el Hierro, one of both is in the wrong location as far as wind is concerned.

      • Rainer says:

        the airport and the windfarm are not at the same place, just the same side of the island. If you look to google map you can see the airport at sea level 31m and if you will find the windfarm find upper storage of GDV and 1km east at 510m high it is. Between airport and Windfarm we have 3 km distance.
        Both places really make sense. At airport and windfarm you have in normal circumstances the wind steady from the sea and not falling down the mountains with gusts.
        We using airport wind data for lack of GDV data. But i am sure that you can put up to airport wind between 20 and 50% to windfarm wind. Tried myself just staying at the rim of the mountains there and had to fight my stand.

  21. Rainer says:

    here you can read the wind at April 9 in time between 00:00 and 1:00 rising from 10m/s to 11m/s. Where you got the 8 to 13m/s from?

      • Kees van der Pool says:

        Either source shows the windspeed at between 8 and 9 m/s for the second half of April 9. The maximum expected wind output would be 4.5MW and the graph shows a sturdy, curtailed 6.4MW. Therefore, the windspeed must be quite a bit higher at the array, which is to be expected. This has all sorts of ramifications for the ‘cut out’ point of the Enercons, the re-configuration of the blade angles and direction adjustments.

        • Rainer says:

          your writing of time:
          “0000 April 9 to 0040”
          Did understand from 00:00 till 00:40.
          Just a misunderstanding…
          I agree absolute that at the wind have more force at the E-70 place than at the airport.

          • Kees van der Pool says:

            No problem – good feedback re: your answer to nukie about maintaining balance at the windfarm. Do you really feel it might be double the airport numbers?

        • Kees and Rainer: The only hard data we have on windspeeds is the XY plot I showed upthread. With this plot you get the best fit between GdV and the airport when you add about 1m/s to the airport wind speeds, which shifts the black dots one grid line to the right.

          Another point you have to consider is wind direction. The GdV wind turbines are aligned east-west along a ridgetop, and I’ve noted (but not documented – too much work) a decrease in output when the wind has a significant east or west component.

          • Rainer says:

            Roger, Kees,
            think your estimate, Roger, of the situation is in basic correct.
            The wind force felt by me is NOT the m/s. As far i know the energy in the wind goes up by the third potency in base of wind speed. Sure the rim works like a “spoiler” only with the correct direction of the wind.

  22. Rainer says:

    2016-04-11 10:00 sudden demand change:
    REE-data demand:
    016-04-11 09:50 5.3
    2016-04-11 10:00 3.8
    2016-04-11 10:10 5.1
    2016-04-11 10:20 4.7
    details following……

  23. Rainer says:

    This time the frequency went down in 1s for 1.5 MW.
    We had a short breakdown here.
    REE data shows:
    Demand: Real Prevista Programada
    2016-04-11 09:30 5.5 5.2 5.3
    2016-04-11 09:40 5.3 5.2 5.3
    2016-04-11 09:50 5.3 5.3 5.3
    2016-04-11 10:00 3.8 5.4 5.5
    2016-04-11 10:10 5.1 5.5 5.5
    2016-04-11 10:20 4.7 5.5 5.5
    2016-04-11 10:30 5 5.6 5.5
    2016-04-11 10:40 4.6 5.6 5.5
    2016-04-11 10:50 4.8 5.5 5.5
    2016-04-11 11:00 4.7 5.4 5.2
    2016-04-11 11:10 5.1 5.3 5.2
    2016-04-11 11:20 5.2 5.2 5.2
    2016-04-11 11:30 5.1 5.1 5.2

    looks like a diesel have broken down
    The frequency meter was connected to a laptop, that worked like a USV.
    The routers in our house have been started again at 10:06……
    The event in detail:

  24. nukie says:

    Diesel went down from 4.6MW to 2.3MW. Second trip of a Diesel Generator we se so far. Hydro jumped in to compensate, as it seems the line rainer is connected disconected, maybe this failure caused the Diesel to trip. So it is not sure if the 47.7 Hz is the real frequency, or the line breaking down. Due to this no frequency data of the failure time itself, just a increased frequency after reconnection of the line, which seems reasonable, to have spare capacity for the moment of reconnection.

    Somebdy should improve the 20kV grid of the island. But the faults of the grid show the behaviour of the different systems during grid failure. So far this year Diesel leads 2:0 against Wind and hydro as far as generator trips are concerned.

  25. Rainer says:

    Thanks Nukie for the Hint,
    did not read the graphs correct!
    The meter have a gap of 889 seconds of data.
    It means 889 or more than 14 minutes and 49 sec no frequency at all at the line.
    here a part of the meter data

    only the first two columns are relevant

    Timestamp in sec f Hz

    1460364824.141 50.01538 95 11
    1460364825.141 49.98942 82 11
    1460364826.142 49.94195 81 11
    1460364827.147 49.92151 40 11
    1460364828.184 47.68942 6 11 —————————- meter gap starts
    1460365717.841 49.95131 19 3 —————————- meter Gap ends
    1460365718.844 49.83599 71 3
    1460365719.847 49.85111 37 3
    1460365720.846 50.06493 35 3

    So no diesel failure but a line failure and the diesel adapted to the situation???

  26. Greg Kaan says:

    Sorry Kees but the wasn’t the whole purpose of the paper was to model the response of the hydro system when coping with fluctations in the grid, given the long pen stock required for pumped hydro?

    The issue I have with simplifying the rest of the grid as a resistive load since the turbines will interact with the pumps in a non-linear manner. I think this is the case because each of the turbines react independently to the grid demand changes (based on what I have read about the Enercon Inertia Emulation). As each turbine reacts, there will be some overshoot in response to the changes in active and reactive loading if the change is less than the sum of the turbines response. Oscillation of this sort is likely to have a greater effect on the DCP tuning criteria as it is inherently more responsive (the whole point of using this vs the DRP criteria).

    There are a lot of ifs and mays in my statements but a assuming linear response with a system with multiple feedback pathways is similarly speculative. IMHO

    • Greg Kaan says:

      My apologies, everyone. This was in response to Kees’s comment http://euanmearns.com/a-preliminary-reservoir-balance-calculation-for-el-hierro/#comment-18061
      I don’t know how it ended up down here

    • Kees van der Pool says:

      Greg, I’m struggling with the following:

      “The issue I have with simplifying the rest of the grid as a resistive load since the turbines will interact with the pumps in a non-linear manner”.
      I have read through the paper a couple of times as well as through a bunch of others on the same subject (there are lots of them). The paper does not mention ‘pumps’ once.
      Its only about adjusting the power to the Peltons which are coupled by means of a shaft to synchronous generators. No pumps.

      Regarding interaction between the pumps and the Enercons: I have no idea how they are programmed but one thing is absolutely certain: they don’t participate in the inertia scheme or in the primary frequency control which is strictly the job of the diesel plant and/or the Peltons.

      To the grid, the Enercons look like an inverter, a bunch of IGBTs and whatnot that synthesize a three phase sine and run with the grid frequency. It would be chaos if they did their own thing regarding frequency. I think there is a trip point at 48.5Hz where the would act reactively for a short while but even that is not certain – it all depends how they are programmed and it would be, short of getting the lowdown from Enercon, speculation.


      • Rainer says:

        stiill learning power technics.
        please: IGBTs ???????

        • Kees van der Pool says:

          Hi Rainer,
          This link is for a solar installation but the DC link part is the same as for the Enercons: http://solar.smps.us/sinewave-inverters.html.
          The Enercon generator is a direct drive, permanent magnet synchronous generator where the three phase AC output of a few Hz is rectified and fed into the central capacitor that feeds the inverter. The three phase AC inverter output is filtered by a transformer and /or a filter array (sometimes ‘active’) before it is fed into the grid. This filtering is crucial as the IGBTs are ‘noisy animals’, very rapidly switching large loads and therefore generating lots of harmonics.

          Assembling these noisy outputs into large windfarms is not always easy, as illustrated by BARD 1 offshore, a 400 MW array of 80 ea. 5 MW windmills. Right after inauguration back in August 2014, things went wrong (in German, I cannot find the English version):
          This is still not fixed, as per the ad-hock ticker of the EEX website:
          The problem seems to be excessive harmonics which leads to reduced output and an occasional HVDC connection outage to the mainland.


          • Rainer says:

            Thank you Kees,
            Think got it.
            will become a rough electrician in my old days. Used to work with mA or micro Amps and digital and not analog….

          • nukie says:

            It wre mainly problems with Bards self developed HVDC-Inverters, as it looks like.
            The bard 5.0 Turbine in this park has no synchronus generator, but a double fed asynchronus turbine, it is a Generation 3 Design. In this design only a part of power comes from a inverter: http://www.4coffshore.com/windfarms/turbine-bard-bard-5.0-tid4.html

          • Kees van der Pool says:

            Well, that proves the point: ‘This filtering is crucial as the IGBTs are ‘noisy animals’, very rapidly switching large loads and therefore generating lots of harmonics”. It was obviously not done ‘right’ at Bard 1 for whatever reason.
            They have the added challenge of combining the 80 ea. outputs, and funneling them into a high voltage DC link.
            Do you have any further information on causes and the way they could be solved? Last October they declared all problems were taken care of but the March BorWin HGÜ failure makes one wonder.

          • nukie says:

            Here some data about the “noisy beast”: http://publications.lib.chalmers.se/records/fulltext/157015.pdf 1-1,5% THD. Below 1% was considered “high fidelity” according DIN 45500 for radio amplifiers.

            For Synchronus generators I do not find measured data, only the requirement that THD shold be below 8% for 95% of time.

      • Greg Kaan says:

        one thing is absolutely certain: they don’t participate in the inertia scheme or in the primary frequency control which is strictly the job of the diesel plant and/or the Peltons

        Kees, do you have any reference for this staement?

  27. jacobress says:

    off topic:
    Have a look at Tasmania’s all green energy system

  28. Rainer says:

    Field-studies-2016-04-13 ca 14:20
    All pictures now also with seconds in the name


    lower storage:
    storage nearly empty…. Roger please verify with your metering
    Peltons working
    Pumps NOT working
    Found a water meter, but do not know if it belongs to GDV or the Diesel Plant.
    Made some pictures to document that it is working hard.
    Can not interprete, need water meter knowledge from others.

    Upper storage:
    storage nearly empty…. Roger please verify with your metering
    water meter did not move at all
    took a short mov-file to show moving E-70 not producing at all at very low m/s.
    Took also pictures of the valve house down of upper level -just from other side of the street where the water meter upper storage is located

  29. Rainer says:

    Example of “Newspaper” in El Hierro.
    Last Info: just switched to 100% renovables. Never switched off……..

    Everybody can leave comments there

  30. I finally found it – frequency limits for the El Hierro grid as per the Spanish Grid Code for small and isolated systems (SEIE):

    Under extreme circumstances, for t < 5 minutes:

    Maximum 50.25
    Minimum 49.75

    From table 2B in

    According to Rainer’s data the El Hierro grid generally operates well within these limits

  31. Rainer says:

    GDV still do not answer my questions and the online paper El Hierro Bimbache do not publish my comments.
    The responsible people are crawling slowly in direction of reality. Will take some time to get there still.

    El Hierro Bimbache public other comments and an other article also.

    Looks more or less similar like the last press paper of GDV:

    GDV now do not talk about 100% but about 70% to 80%.

    “Con las condiciones actuales de la Central, como se ha expresado en numerosas oportunidades, técnicamente podemos abastecer entre el 70 y el 80% de la demanda eléctrica anual, cifra que alcanzaremos con la buena evolución que revela el proyecto”, afirma Sánchez, quien además recuerda que la página web del operador de sistemas, Red Eléctrica de España, permite hacer un seguimiento a tiempo real de la estructura de generación de la Isla, lo que desde Gorona del Viento continuamente se invita a seguir de manera que los herreños puedan compartir con la empresa la satisfacción del buen hacer de la Central, actual esperanza para las renovables entre los expertos que trabajan en pro de las energías limpias.”

    “With the current conditions of the Central, as has been expressed on numerous occasions, we can technically provide between 70 and 80% of the annual electricity demand, a figure that will reach with the good performance that reveals the project,” says Sanchez, who further recalls that the website of the system operator, Red Electrica of Spain, allows you to track real-time generation structure of the island, which from Gorona the wind continually invited to follow so that islanders can share company with satisfaction the good work of the Central, now hope for renewable among experts working for clean energy.

    Also GDV still refuse to give online data:
    No can imagine that it is forbidden to give really data at the public.

    “Cabe destacar que, si bien el Cabildo es el socio mayoritario de la empresa, se trata de una sociedad anónima cuyo objeto es la generación de energía y que se rige por los criterios mercantiles de cualquier sociedad, que estipula el grado de información pública inscrito en el Registro Mercantil.”

    Notably, although the council is the majority shareholder of the company, is a corporation whose purpose is the generation of energy and governed by commercial criteria of any society, which provides the degree of public information inscribed on the Commercial Register.

    • Rainer: Near the bottom of the post they refer to a graphic showing that GdV has achieved only 31.1% renewable output since startup. I’m sure this is one of our graphics but it shows up only as a little box in the version you sent me. Do you have a copy of the article that shows the graphic?

  32. nukie says:

    @ Kees, I did not find any details how ABB solved the problems with the BARD-Design of the HVDC-Inverter. But since it was not their own design they had to repair, they had some work with it.
    This paper also gives no explanation: http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7268916 it just tells that parametes of the VSC can influence damping of the possible oscillations between VSC, cables, and generators within a wind farm.

    • Kees van der Pool says:

      Thanks for the interesting link, nukie. I don’t think the problems are solved yet and maybe ABB should tear Borwin apart and start all over with their own stuff. For whatever its worth, the last I read was on Pierre Gosselin’s blog by DirkH (who else?) who described successively shutting down each windmill to re-write the (individual?) code for every single one of them.
      There is remarkably little information on the web regarding the progress, such as it is, of BARD 1.

  33. Rainer says:

    Diesel off: 2016-04-26 07:20
    Diesel on: 2016-04-26 13:20
    +6h=183h or 2,09% of a year

  34. Rainer says:

    Diesel off: 2016-05-02 00:00

  35. Pingback: El Hierro, March/April 2016 update: | Energy Matters

  36. Rainer says:

    2016-05-02 12:00 Diesel on
    +12h=195h or 2,23% of 8760h, years hour
    more info thanks to Roger:

Comments are closed.