The Gorona del Viento (GdV) plant on the Canary Island of El Hierro is a flagship project designed ultimately to provide the island with 100% renewable electricity and to demonstrate that hybrid wind/pumped hydro systems can be used to generate 100% renewable electricity in other parts of the world. GdV comprises a wind park with 11.5 MW capacity and a pumped hydro storage plant with 11.3MW capacity, installed at a total cost of €84 million. This is the fifth in a series of operational updates that began in September last year. Details on GdV plant layout, operation and capacities are given in the September update.
The low-wind conditions that dominated during the last four months of 2015 continued into January 2016, leading to only 22% renewables generation in that month. Much of February, however, was characterized by strong winds, and combined with the 100% renewables tests that were performed this resulted in renewable energy supplying 54% of El Hierro’s grid demand in that month, exceeding the 52% achieved in June/July 2015.
El Hierro monthly grid statistics are summarized in the Table below. Note that wind and hydro generation sent to the grid are now combined into a single “renewables” category. The reasons for this are discussed later:
Figure 1 plots average daily diesel and renewables (wind + hydro) generation since full operations began on June 27, 2015:
Figure 1: Average daily generation since project startup
Figure 2 plots the 10-minute Red Eléctrica de España (REE) grid data for January and February 2016. Features of interest include the four periods of 100% renewables generation (16 hours between 0540 and 2140 on January 31, 41 hours between 0100 on February 14 and 1740 on February 15, 24 hours between 0110 on February 20 and 0120 on February 21 and 35 hours between 0020 on February 28 and 1050 on February 29) and the two grid “crashes” around 5 am on February 18 and between 3 and 5 am on February 19, which were probably triggered by high winds and rain:
Figure 2: January and February 2016 generation, 10-minute REE data
In previous updates I’ve found that I repeat myself a lot, so in this one I will discuss the January and February results relative to some of the questions and observations that arose in the recent “16 hour” and “41 hour” posts, which so far have generated almost 300 comments between them:
1. Segregating wind and hydro
The Red Eléctrica de España (REE) 10-minute grid data (accessible via the “El Hierro Live Grid” sidebar) provide three numbers for the El Hierro grid – diesel, wind and hydro. REE has confimed that the diesel and wind numbers show actual generation from the Llanos Blancos diesel plant and the GdV wind farm. The hydro number, however, shows net generation from the GdV hydro plant less power consumed in the pumping plant and doesn’t distinguish between the two. Moreover, according to REE the pumping station and the hydro plant are both often working at the same time. As a result the hydro numbers are ambiguous. A value of minus 1MW could, for example, represent 1MW of pumping and no hydro generation, or 1.5MW of pumping and 0.5MW of hydro generation, or 2MW of pumping and 1MW of hydro generation. There’s no way of knowing which.
The graphs in previous posts that segregated wind and hydro into green and blue bars assumed that the negative hydro values represented pumping and the positive values hydro generation. This assumption was broadly in line with GdV’s operating scheme, which is to send as much wind power as possible directly to the grid to avoid the 40% energy losses incurred in the hydro circuit. But when grid stability is factored into the equation GdV may well have a preference for hydro over wind despite these losses, and consequently it’s likely that previous analyses will have overestimated wind sent to the El Hierro grid and underestimated hydro. For this reason wind and hydro are now combined into a single “renewables” category, although how much of the renewable energy is wind and how much hydro makes no difference to the total amount of renewable energy sent to the grid.
2. “Windy Sunday” tests and grid stability
100% renewables tests have been conducted on four of five Sundays since the wind began to pick up in late January (Figure 3), with Sunday selected presumably because it’s the day a grid outage would have the least impact. The tests have continued either until the wind dropped, which was the case during the January 31 and February 14/15 test, or until the diesels were restarted even though wind generation still exceeded demand, which was the case with the February 15/16 and 28/29 tests. This has prompted the question “why weren’t the February 15/16 and 28/29 tests allowed to continue?”
Well, REE hasn’t told me why, but clearly there was something about the tests which suggested to REE that GdV wind still isn’t quite ready for prime time. Here it’s important to recognize that the stability of the El Hierro grid is REE’s responsibility, and that in deciding what the grid can and can’t accept REE has to meet the Spanish codes for island grids set forth in the Boletín Oficial del Estado 2006 – which according to Merino et al “are among the most demanding international codes currently in force for wind power integration in electrical systems” – and also has to conform with various other Spanish and EU directives governing the management of grids. REE also isn’t performing the tests to see how long 100% renewables generation can be stretched out before something breaks. The purpose is to evaluate the grid’s ability to recover promptly from a major fault, such as a busbar short-circuit, when 100% of the generation is coming from renewables. And the fact that the El Hierro grid took 24 hours to recover from the unscheduled outages that began around 5 am on February 18 suggest that it isn’t very good.
The way in which GdV’s output is being handled, particularly the curtailment of wind at around 7MW (discussed later), has been described as “insane” by some commenters, but there’s really nothing insane about it. What GdV has done is adopt effectively the same solutions to the wind instability problem as has King Island, Tasmania, a closely comparable project with battery storage instead of hydro. So if GdV is insane Hydro Tasmania is too. But this is nevertheless an important issue because it tells us what other high-penetration renewables systems will probably have to do to ensure grid stability. The two key ingredients are:
- A dynamic resistor to control frequency fluctuations caused by variable wind output. King Island’s resistor does this by converting excess wind to heat. GdV does it by sending excess wind to pumping.
- A means of adding inertia to the system. King Island does this with a flywheel. GdV does it by keeping at least three of the 2.83MW Pelton hydro turbines spinning.
A dynamic resistor, however, inevitably involves some wastage of wind power. An example of wastage at King Island is shown in Figure 3 below, reproduced from the King Island Renewable Energy Integration Project document . Maybe 15-20% of the available wind energy is curtailed:
Figure 3: Generation mix for King Island, Tasmania, showing wind wastage
As shown in Figure 4 wastage at GdV is higher – an eyeball estimate suggests about 30% during the 41-hour period of 100% renewables generation during February 15 and 16 even when wind farm output is curtailed at around 7MW, about which more later:
Figure 4: Generation during February 14/15 renewables test, 10 minute REE data
But what we are left with is a case where grid demand has been met with 100% renewables for 41 hours simply by switching wind output between the grid and pumping and without having to adjust generation. At the cost of the wastage of wind power that couldn’t have been used anyway GdV’s intermittent wind output has been converted into dispatchable power. And if wind (or solar) energy can’t be made dispatchable there is no way a high-penetration renewables project will ever succeed.
Figure 4 also reveals some details of the test. Once every 13 or 14 hours wind farm output was abruptly reduced for an hour or two presumably to see how the grid responded, and during the second and third tests wind was briefly “spiked” back on again towards the end of the test. The responses to the first and second tests were evidently acceptable but the diesels were turned back on immediately after the third. This may have been because the response in this case was unacceptable or simply because REE had obtained all the information it had planned on getting from the test.
Another feature of Figure 4 is the flat-topped wind output, which shows that wind was deliberately curtailed at around the 7MW level. Why, asked some commenters, curtail wind at this level? GdV has 11.5MW of installed wind capacity, why not use all of it? In this particular case the reason is obvious – all the additional wind power would be wasted, so there’s no point generating it, and the unconstrained wind surges would probably have a destabilizing influence anyway. The question is really why GdV can use only some of the wind power it’s capable of generating, and we’ll get to that later.
3. Irrigation and water supply impacts
Because GdV and the El Hierro Island water council have not responded to requests for data we have had to resort to detective work to estimate two important numbers – how much desalinated water has been delivered to the GdV lower reservoir and how much has been extracted from the upper reservoir for irrigation or other uses.
There is general agreement that there are three desalination plants on El Hierro:
- El Golfo: One plant, capacity 4,000 cu m/day
- Tamaduste (also known as El Cangrejo): Two plants each of 1,200 cu m/day capacity, total 2,400 cu m/day
- La Restinga: Two plants of 1,200 and 500 cu m/day capacity, total 1,700 cu m/day
These plants have a total production capacity of 8,100 cu m/day assuming that all operate continously.
Figure 5 is a map of El Hierro showing the locations of the three plants relative to the GdV reservoirs and the island water pipeline network (for a closer look at the pipeline network this map from the Consejo Insular de Aguas de El Hierro can be enlarged to almost any scale). The tabulation of plant specifications at the bottom of the Figure is from the Official List of Canary Island Desalination Plants, dated 2013. Note that the expansion of the El Golfo plant from 1,350 to 4,000 cu m/day in 2012 isn’t shown on the list.
Figure 5: El Hierro’s desalination plants and water pipeline network. Red pipelines are pumped and blue pipelines gravity-fed.
Let’s now see how these three plants are positioned to deliver water to GdV:
First El Golfo. The El Golfo area is the only respectably flat piece of land of any size on El Hierro and it’s here that effectively all of the irrigated crops (mostly bananas) are grown. El Golfo’s water has historically come from water wells but now it will come from the El Golfo plant to avoid further groundwater depletion. The plan is to use all of the plant’s output in the El Golfo area.
But let’s assume that some of El Golfo’s water does get sent to GdV. How does it get there? By a tortuous pipeline route that requires pumping up to an elevation of 770m close to the island capital of Valverde. And shortly after this the water arrives at a point immediately above the upper reservoir, whereupon the logical solution if the upper reservoir is to be used for fresh water storage is to drain it down into to the upper reservoir. It certainly makes no sense to let the water drain all the way down to the lower reservoir and then pump it back up again.
But why send water to GdV at all when it can be used locally?
La Restinga has the same problem, only worse. Any water sent from here to GdV will have to negotiate about 20km of winding pipeline, will have to be pumped up to an elevation of 1,130m and will eventually arrive at the same point above the upper reservoir as the El Golfo water did.
Tamaduste is the only desaladora that is realistically positioned to deliver water to GdV, but at a rate of only 2,400 cu m/day even if it operates at full capacity for 100% of the time, which it probably doesn’t, and if there are no other demands on its production, which there almost certainly are. But we will nevertheless assume that Tamaduste ran flat out during July and August 2015 and that all of its output was delivered to the GdV lower reservoir. This gives 62 days times 2,400 cu m/day for a total of 148,800 cubic meters delivered.
And how much water was pumped from the lower reservoir to the upper reservoir during July and August? The REE “hydro” grid values show that 2,743 MWh was consumed in pumping during these months, enough to pump over a million cubic meters of water from the lower to the upper reservoir.
So we have 150,000 cubic meters entering the lower reservoir in July and August and over a million cubic meters leaving it. The only possible explanation for this imbalance is that the water pumped from the lower to the upper reservoir was allowed to flow straight back down again. Why was so much water pumped uphill in the first place? Because GdV has been using pumping as a dynamic resistor during high-wind periods ever since the project went into full operation in late June, 2015. As shown in Figure 6, the July 2015 generation mix was in fact very similar to February 2016 generation mix (Figure 2) except that no 100% renewables tests were performed during the month:
Figure 6: July 2015 generation, 10-minute REE data
And if almost all the water pumped uphill came straight back down again we can reasonably assume that very little water was extracted from the upper reservoir for irrigation or other purposes.
4. Why GdV will never achieve full-time 100% renewables generation.
One reason GdV will never achieve 100% renewables generation as currently set up is the use of the GdV reservoirs as a dynamic resistor, which as discussed above will lead to the curtailment of a significant amount of wind generation and probably leave the existing 11.5MW wind farm incapable of generating enough electricity to fill annual island demand. It’s important to note here that a dynamic resistor was not part of the original operating plan, which assumed that grid stability at high levels of wind penetration could be achieved simply by keeping three Peltons as a spinning reserve (see Merino et al, link above). But it’s clearly a requirement if the system is to function, and GdV has to live with it, as does King Island 11,000 miles away on the other side of the Earth. And the stability problem at GdV still hasn’t been solved, although it’s clearly moving closer to solution.
But that isn’t the main obstacle.
We will now assume that grid stability isn’t a problem and that all the wind power the 11.5MW wind farm is capable of generating can be sent directly to the grid or to storage. Assuming an optimistic 40% capacity factor this gives a total wind farm output of 40GWh/year, just about enough to cover El Hierro’s annual demand. But there will of course be periods when wind generation exceeds demand and the surplus wind must be stored for reuse, and other periods when wind generation is lower than demand and the energy in storage must be released to cover the deficit. And if GdV’s reservoirs don’t have the capacity to handle these storage and release requirements, well, it’s back to diesel power.
Now we will look at a specific example – October 2015, the worst wind month since operations began, although December wasn’t much better. The GdV reservoirs have an energy storage capacity of about 250MWh (limited by the 150,000 cu m capacity of the lower reservoir) and I start by assuming that they are fully charged at 0000 hours on October 1, although as a practical matter they won’t be because the second half of September was windless too. How long does the storage last?
Figure 7 shows that it runs out at midnight on October 4, leaving a 3,000MWh deficit between demand and wind generation over the rest of the month that would have to be filled with diesel generation.
Figure 7: Duration of energy storage, September 2015, current design
Is there anything that can be done to beef up the storage capacity of the reservoirs? An obvious option would be to use the ocean as the lower reservoir, which would increase storage from 250MWh to about 630MWh, limited by the 380,000 cu m capacity of the upper reservoir. But this does not greatly improve matters. Storage runs out around 6pm on October 7 and we are still left with a demand shortfall of 2,600MWh over the rest of the month:
Figure 8: Duration of energy storage, September 2015, storage increased to 630MWh by using the ocean as the lower reservoir
The only other practicable option would be to increase the size of the wind farm. Figure 9 shows the impacts of doubling its capacity from 11.5MW to 23MW while keeping the ocean as the lower reservoir. They are again not very large. The storage now runs out at midnight on October 8 and we are still left with a 2,200MWh demand shortfall over the remaining 23 days of the month:
Figure 9: Duration of energy storage, September 2015, storage increased to 630MWh by using the ocean as the lower reservoir, wind farm capacity doubled to 23 MW
And if we were to allow for the fact that low-wind conditions were common between mid-September 2015 and late January 2016 the demand shortfalls would increase substantially.
The same thing happens in reverse, although in this case there’s no way of doing an analysis because we don’t know what the wind generation profile would have looked like without curtailment. But there were certainly occasions in July and August when surplus wind generation could have charged the reservoirs from empty within a few days, leaving no room for any surplus generation that came after.
Clearly the storage capacity of the GdV reservoirs is at least an order of magnitude too small to allow the project to supply El Hierro with 100% year-round renewable energy or anything close to it. The question is, why wasn’t this recognized earlier? Well, it almost certainly was. But the “innovative” GdV reservoirs had received such overwhelming international acclaim that it would have been a brave person indeed who dared to tell the project’s backers that their baby was ugly. As a result GdV is still widely touted as a 24/365 100% renewables project, or something very close to it, even though it has no chance of becoming one. Here’s the official line from Gorona del Viento:
The operation’s philosophy is based on supplying the electrical demand of the island with renewable sources, thus guaranteeing the stability of the electrical network; the diesel engine plant will only operate in exceptional/emergency cases, when there is not enough wind (or) water to fill energy demand.
How often do exceptional/emergency conditions occur? Five percent of the time? Two percent? One? Sorry, but the diesel engine plant is going to operate a lot more often than that. Based on results to date Hubert Flocard and I in fact estimate that diesel generation will be supplying island demand for about 50% of the time.
The fact of the matter is that GdV’s performance will be hostage to the wind for the rest of its operating life. During high-wind periods such as occurred in July, August and early September last year and in the second half of February this year we can expect that the plant will ultimately be capable of supplying 100% of El Hierro’s electricity with renewable energy for extended periods. But during low-wind periods such as occurred between mid-September last year and mid-February this year diesel generation will have to provide the bulk of the island’s needs. The basic problem is that GdV has every conceivable kind of smart gadget in its control room except for the one it really needs – a gadget that makes the wind blow to order.