Windmills increase fossil fuel consumption and CO2 emissions 

Abstract

First we describe the models presently used by others to calculate fuel saving and reduction of CO₂ emission through windparks. These models are incomplete. Neglected factors deminish the calculated savings.

Using wind data of a normal windy day in the Netherlands it will be shown that windparks of various size cause extra fuel consumption instead of fuel saving, when compared to electricity production with modern gas turbines only. We demonstrate that such losses occur.

Factors taken into account are: low thermal efficiency at low power; cycling of back up generators; energy needed to build and to install wind turbines; energy needed for cabling and net adaptation; increase of fuel consumption through partial replacement of efficient generators by low-efficient, fast reacting OCGTs.

1. Introduction

Several countries invest heavily in the construction of windmills in order to save fossil fuel and to reduce CO₂ emission. The wind comes free, the mills do not pollute and there is no need to burn fossil fuel. However, this simple notion defended by staunch supporters of windturbines, has been criticized by critical analysts, e.g. refs 4, 5, 6, 8, 10, 11, 12.

Wind does not blow according to demand of electricity users. Sometimes there is no wind or little wind and sometimes there is a lot. It would be no problem if there was an economic way to store electricity and to use it from that storage whenever needed. Unfortunately we do not have such a storage. Batteries have little capacity and they are much too expensive. There are other possibillities but none of them comes near to anything that is economically feasible. The only exception is hydro power, i.e. lakes in mountains, that can be pumped full if there is an electricity surplus and emptied when the power is needed. Unfortunately there are no mountains in the Netherlands. (Also many other countries that do have them, do not have sufficient place there for such storage lakes.) So the current practice is to have windparks operate in connection to conventional powerplants. These generators step in when the wind fails and they can be switched off, or their output is reduced, if the wind blows. Thus, when considering wind power, one must do that normally in connection with ‘back up’ conventional systems. That is why the wind influences from minute to minute the performance of the conventional generators.

A handicap prohibiting the settlement of the dispute is the absence in the public domain of factual data of the different producing units. So the the arguments are mostly about model computations. There are exceptions. In the USA a BENTEK study used real emission data of power plants in Texas and Colorado. They became available due to the freedom of information act. Its conclusion was: wind has no visible influence on fuel consumption for electricity production and the emission of CO₂ in the atmosphere is not reduced¹³.

This shocking result did not convince decision makers. At least not in Europe. The negative result was attributed to a difference in fuel mix. Coal-, oil-, gas- and nuclear heated generators behave differently. So what might be true there, does not mean that it holds true for us.

In August 2011 Fred Udo analysed the data put on the internet by EirGrid, the grid operator in Ireland. His web page article was termed by colleagues abroad ‘The smoking gun of the windmill fraud’. He showed that the substantial wind contribution in the Irish republic caused such a small saving of fuel and a corresponding small reduction of CO₂ emission, that it shatters the whole economy of the wind policy. He also was able to show that more wind penetration caused an increase of CO₂ emission⁸.

The real situation, however, is even worse. The way EirGrid derives its data on CO₂ emission does not correspond with what is actually happening in fossil fired power plants. More over, the Irish data do not enclose some serious other factors that deteriorate the fuel saving aimed at. An indication could be, that the overall CO₂ emission in Ireland is 20% higher than the emission calculated in the EirGrid tables, as Udo showed. (His source: ref. 14. A difference of 3% might be due to import of electricity. Transport losses have been accounted for.)

In this present study we shall explain what is wrong. On the basis of existing data and new information on the behavior of conventional generators when they are cycling – i.e. ramping up and down in order to compensate for the variations in wind power – we shall show how much worse the influence of adding wind electricity to the grid really is.

2. The old model.

During the early days of modern wind turbines the argument was simple and appealing. Every kWh electric energy generated by wind replaces a kWh produced by a conventional power plant. As a result the fuel needed to produce it, is saved and the CO₂ that would be produced is not released in the earth’s atmosphere.

Different generators have different thermal efficiencies. And the CO₂ production is different for gas, coal and other fuels. Some basic data for certain generators and fuel used in the Netherlands are listed in table 1. The coal fired unit in the table is the most efficient one currently under construction. Others presently in operation do not have efficiencies better than 0,44 or 0,42. The data about CCGTs should be read as of the best units running at the moment. The newest OCGTs may under optimal conditions reach an efficiency of 0,36. But there is quite a number of older ones that will remain in operation till after 2020.

Table 1.

Thermal efficiency η of different generators1
coal fired steam enhanced gas turbine, CCGT open cycle gas turbine, OCGT nuclear	0,455 0,59 0,32 0,377
latent heat2
Gas [J/m3] coal [J/kg] Uranium [kWh/kg]	32 × 106 29 × 106 7,4 × 106
CO2 emission when burning3
Gas [kg CO2/m3] Coal [kg CO2/kg] Uranium	2,5 2,6 nihil

The old model then tells: for every kWh wind electricity we save fuel and gas as is summarised in table 2.

Table 2.
Savings according to the old model.

Type generator	Per kWh conventional	Per kWh CO2 [kg]
Coal fired CCGT OCGT Nuclear	0,273 kg coal 0,191 m3 gas 0,352 m3 gas 0,358 mg Uranium	0,71 0,48 0,88 nihil

These figures have opened the market for the large size windmill introduction. Governments and the public became convinced. Wind offered a possibillity to offset the thread of climate change and depletion of fossil stocks. Even today the same numbers are often used in public debates, sometimes disguised in terms of ‘so many windmills are capable to provide for the electricity needs of so many households’.

Critics pointed to flaws in the assumption. There are several reasons why these figures are wrong. This lead to a new model, which is now accepted e.g. by the Dutch government. (Also the EirGrid uses this model in order to calculate the CO₂ emission on the basis of the amount of electricity produced by the different conventional power plants during their operation in co-operation with the windparks.) We shall therefore call it the current model.

3. The current model.

The current model acknowledges variation of the thermal efficiency of the generators. A generator is designed for a certain optimal output. If one lowers the temperature, i.e. feeds in less fuel, the electricity output does not deminish proportionally. Every conventional generator has its own ‘heat rate curve’ describing how its efficiency, η, depends on power output. With increasing wind electricity penetration conventional power generation has to be reduced and the efficiency of the units becomes less. This reduces the savings calculated with the old model. The results are presented in table 3. For data & algorithm see Appendix.

Table 3.
Comparison of savings in fuel and CO₂ emission between the old model
and the current model with resp. 20%, 40% and 60% less than
‘design power’ of the back up conventional plants.
We left out the ones not relevant: “n.a.”

savings					relative setback
old model all	current model
	Coal fired	CCGT	OCGT	nuclear	CCGT	OCGT
0% 20% 40% 60%	0% 19,1% 36,5% n.a.	0% 17,2% 35,6% 51,8%	0% 14,7% 28,9% 44,3%	0% 17,8% n.a. n.a.	n.a. 14,0% 10,9% 13,6%	n.a. 26,7% 27,8% 26,1%

The current model shows saving under the given circumstances. The Netherlands government assured parliament that the previously assumed savings had to be reduced by at most 10%. When we look at the figures in table 3, we see that it was slightly overplaying its hand. Our calculations show a relative savings reduction exceeding this for the most relevant generator types, CCGT & OCGT. OCGTs ought to be used as little as possible in view of their low efficiency. They are only needed when rapid power changes are required.

(Coal and nuclear plants are almost irrelevant in this respect as they cannot be ramped up and down sufficiently fast to follow wind variations. Nuclear plants do not produce CO₂ anyway and their fuel is virtually inexhaustible.)

4. Errors in the current model.

Unfortunately the current model does not represent what is going on in a power plant. It neglects completely other factors that reduce the supposed fuel and emission savings. We shall first list the important factors that influence the fuel consumption and the savings. After that we discuss them and show their implications.

1. Cycling, § 5.

   As we mentioned before, cycling i.e. ramping up and down of conventional generators, differs from running them at less than their designed power in a stationary mode. The latter can be dealt with using the well known ‘heat rate curves’ for that particular type of generator. For cycling there is no public data. If it exists, it is kept secret. The power industry world wide consider it ‘competition sensitive’. We have argued several times that cycling is important because it is inherent to the task of following wind energy variations. It has such a strong impact on the fuel consumption of the plants, that authorities should insist that this data becomes available before they decide on huge subsidies for the wind industry. The argument that generators did cycle also before wind electricity was added because of variations in demand, is irrelevant. The wind variations add up almost to their full extend and they are more frequent and less predictable than demand variations, see for instance figures 1 & 3 below.

   Recently we received some information concerning a fuel flow recording of a coal fired generator during cycling. The generator running stationary for some time at 100% of its optimal capacity reduced its output to 80% and up again to 100%. The whole cycle took place in one hour. The total fuel consumption during that period was 1,2% more than it would have been had the machine continued running at 100%. It was suggested that for a CCGT this outcome should have been 1%⁷. One might wonder whether this measurement is at all representative for the conventional segment? There is good evidence, that it is. A few decennia ago power companies in the Netherlands were owned by public authorities, cities or other regional entities. They were nation wide united in a co-operative association, the SEP. Within that organisation there was a free exchange of information. The SEP regulated the production of the individual plants in such a way that variable costs were minimised. Therefore the individual heat rate curves were precisely known. Please note: these were measured data, not theoretical! It turned out that the actual fuel use of the units doing the regulation and delivering the variable part of the power needed, nation wide, was always some 0,3 – 0,5% higher than that calculated with the heat rate curves. This remarkable difference was attributed to the ‘hysterysis effect’. Variations in demand required the plants to ramp up and down causing this extra fuel consumption. One should take into account that some 30% of the joint producing units took part in this cycling and provided for the extra demand above the permanent load. The demand variation was higly predictable. It consisted more or less of only two major cycles per day and yet 0,3 – 0,5% more fuel for the whole top production. This strenthens our trust in the validity of the figure of the test run.

   In our calculation later on we shall assume this behaviour as a cycling fact¹⁵.

2. Energy costs of construction and installation, § 6

   Windmills are considerable units. They require energy for their constituents, their construction, their foundation and their installation. One of the firms actually doing this type of work figured it out. (See ref. 5. Note 13.) It boils down to an amount of energy equal to the assumed production of the wind turbine during a period of 1½ year.

   This energy investment has to be ‘written off’ during the whole life time of the installation. This according to wind supporters is supposed to be around 25 year. We have seen recently that a whole windpark in the Netherlands with that supposed life time had to be renewed after 12 year. Subsidy regulations applied by the government are based on a write off in 15 year. That is the period we deem realistic.

   We shall incorporate the energy costs factor in our subsequent calculations with a life time of 15 years. To appease the wind fans, we’ll add a line based on 30 year.

3. Energy costs of connection and adaptation to the grid, § 7

   The same as in b must be assumed for the extra cables and the adaptation of the wind generators to the grid. Germany has to construct for instance 2700 km extra high power lines. The Netherlands for that reason was connected by under water cables to Norway and to the UK. The Norwegian connection had already to be renewed partly two years after initial construction. The new to be built off shore windpark in Denmark ‘Gwynt y Môr’ will cost ~ 2 G€. 1,2 G€ of that is required for the wind turbines, 0,8 G€ for the connections etc.

   We shall include in our calculations a similar ‘write off’ for this purpose as for the energy costs in b above.

4. Need for more OCGTs, § 8

   There are two types of gas fired generators fit to co-operate with the wind turbines: CCGTs and OCGTs. CCGTs are beautiful effective machines. Their efficiency might before long reach a thermal efficiency of 60%. However, their ability to ramp up and down is not suited for very rapid variations. It is in the order of ½ hour. But frequent ramping is unlikely because of the damage in terms of wear and tear (see g. below) it causes. OCGTs on the other hand can deal with variations within minutes. But their efficiency is sadly low. It is about 32% while running at design power. The wind variations may sometimes come sudden.

   The centralised grid regulation is to a large extend done on the base of frequency regulation. This requires sophisticated manipulation of the available units. As a consequence units are often condemned to operate on less than their design power with less than optimal fuel efficiency.

   Therefore it is necessary to make more often use of the OCGTs than would be the case without wind power. More use of OCGTs means more fuel. It reduces the savings the wind might give.

   In our calculations we have made a moderate estimation for this factor.

5. Quasi static ramping, § 5 (cycling)

   In the current model it is assumed that there is an instantanious transition in a cycle from one stationary state to the next with different η. In reality there is a transition that takes time. In our calculations we have used a slightly more sophisticated approach. We assume the transitions to take place as a quasi static proces. (The cycle loss is taken into account separately.) This means that at any time during the transition we account for conditions pertaining to those at the power level at that moment. The results as shown in table 3 are not significantly altered. In a more frequently occurring ramping up and down in which the transiton time becomes more important with respect to the time in which the generator is in a stationary state, there is a difference.

6. Self consumption of electricity

   Windmills do not only produce electricity, they also use it. Electricity is needed to start them, and to heat some of their parts. The power regulation electronics consume electricity all the time. It is not known, whether the actual production data provided by the national statistics bureau, CBS, are nett figures. We suppose that the turbines, while running, provide for their own needs. But when they are not producing, that cannot be the case.

   For the time being and by lack of information we have not included this element in our calculations.

7. Extra wear and tear

   Life time and maintenance of conventional plants depend largely on the ramping activity. More than on the number of stationary running hours. Ramping is a fact of life in the electricity business because of variations in demand. However, the connection with wind power adds extra to the normal, that is according to demand variation, cycling routine. Also the wind varations are often less predictable. This issue was reason for the government to ask for a special assessment. The report¹ of a research group at Delft University of Technology came out in April 2009. It contains serious warnings about this phenomenon. In the USA there are firms active, which make their business by consulting power producers about more efficient ways to deal with ramping in order to save on extra fuel costs and to protect their costly equipment against faster wear and tear than what is absolutely necessary. The extra maintenance and life time shortening must have consequences in terms of energy costs.

   We have to omit this factor in our calculations by lack of sufficient information.

8. Spinning reserve

   In actual situations it happens that conventional units must be turned off because of the wind electricity preference. In such cases normally these units remain spinning idle and thus are using fuel without production of electric energy. Data there about is also not available.

   We also have to omit this factor in our calculations by lack of sufficient information.

-o-o-o-o-o-o-o-

Towards an integral savings assessment of windturbines.

(Details of the calculations can be found in the Appendix.)

5. Cycling.

The biggest CCGT presently in operation has a maximum capacity of 440 MW. In our model we use a hypothetic gas fueled plant with a capacity of 500 MW. In combination with a 100 MW windpark 3% or 15 MW of this is supplied by an OCGT. In that case the remainder has to be supplied by two smaller CCGTs. For a mainly CCGT based plant with a design capacity of 500 MW the cycle properties as described in § 4a implicate that the assumed fuel saving during one hour with a cycle 100% – 80% – 100% and a ramp rate of 12 MW/min actually becomes a loss in stead:

assumed saving	~ 16 400 m3 gas
actual loss	~ 950 m3 gas

The substantial difference is not so surprising; think of a car in town and on an express way. The fuel use of a normal diesel engine while driving at a constant speed of 100 km/h is normally about 50 – 60% of its consumption in a city with continuous speed variation. This happens also with power generators that have to adjust their output continuously following the variations of their wind powered counterparts.

We now consider a region to be served by a windpark in combination with a conventional system. We assume a constant demand of 500 MW. The conventional system, we choose, consists of the most efficient generator units (CCGT), only when necessary assisted by a small fraction of OCGT. In order to cope with lulls in the wind, the conventional power system has a design capacity of 500 MW. For the wind park we shall look at 100, 200 and 300 MW name plate capacity. To approach average conditions, we’ll choose a normal windy day, picking the wind record of Schiphol Airport on August 28, 2011.

Figure 1.

A wind turbine depends for its power on the flow of the wind energy, i.e. it varies with the 3d power of the wind speed, v. If v ≤ 5 kn (= 2,5 m/sec) the turbines do not produce electricity. (Their wings may still be rotating. That is better for the bearings, but there is no output.) At 19 knots they reach their maximum capacity, i.e. P = 100 MW (or 200, or 300 MW).

In between the power must be interpolated by:

P = 0,03644 × (v − 5)3 (or 2x, or 3x)

(1)

as depicted in

Figure 2.

This implies a loss that depends on the wind speed. At that site on August 28 this means a varying wind contribution shown in

Figure 3.

If we would calculate the total power contribution using the old model, on that day the 100 MW wind park would have saved 4,2% of the use of the conventional power plant. Details of the arithmatic can be found in the Appendix. Wind power experts attribute a ‘capacity factor’ of 25% to wind mills in the Netherlands. That is to say with a name plate capacity of 100 MW the average contribution over the year would be 25 MW, which means 5% saving. However, in 2008 the overall capacity factor of the wind turbines in the Netherlands was 22,63%¹⁶. Thus an average wind day in 2008 would have saved 4,5%. We found 4,2% which means that August 28 was just slightly less than an average wind day that year.

But, because of the cycling effect the real result is appreciably different.

The Schiphol record tells us the wind speed every half hour (i). With (1) we find the wind power, P_w,i, and the power of the adjusting CCGT+, P_GT,i.

PGT,i = 500 − Pw,i

(2)

We calculate the fuel consumption during that half hour with the quasi stationary method. That is we split the 30 minutes in a part in which the CCGT+ produces stationary and a part at which the system is ramping from the previous level P_GT,i-1 to P_GT,i. For the first we part we use: η_i and for the second: 0,5 × (η_i-1 + η_i).

For details see Appendix.

We know what cycling does in the case of 100% – 80% – 100% for a 500 MW generator. During a full cycle there are three phases: up, down & stationary. In a cycle during a full hour with 12 MW/min, these last ~8,3 min, ~8,3 min and ~43,4 min. During the 43,4 min there is saving. During the two times 8,3 min there is saving while going down and extra fuel consumption going up. The nett cycle cost in the example is the same whether it happens during an hour or during a half hour as long as the ramp rate remains 12 MW/min. Only the stationary minutes are less.

The net cycle costs will depend on the amplitude of the cycle and to some extend on the power level at which the CCGT operates. Also the duration of cycling depends on the amplitude.

We assume:

1. The net cycle loss does not depend on the power level. (Probably the loss at low power i.e. with high wind penetration increases, because the relative diferences are bigger and the CCGT has a lower efficiency there.)
2. The nett cycle loss is proportional to the amplitude. It is zero if the amplitude = 0 (stationary) and at amplitude = 100 MW the loss is as in the example.
3. During the half hours we only see half cycles. We assume that they require also only half the nett loss. Because the wind speed over a longer period always returns to its earlier value, our CCGT ramps as much up as down, which justifies this assumption.

Now we are able to compute for each half hour the savings of the system. (Quasi stationary saving minus the pertaining cycle loss.) Summing them up and comparing them with the fuel use of our CCGT at full power, we obtain the percentage fuel (and emission) saving over the 21½ hour period.

6. Energy costs of construction and installation.

We use the data of the energy costs for construction and installation of the research department of Volker Wessels Stevin, a major installer of windturbines⁵: 1,5 year windmill production to recover the needed energy. If we then assume the life time of a windmill to be 15 year, it means that 10% of its production must be deducted to compensate for the earlier loss. We shall also do our calculation for a life time of 30 years, meaning a 5% deduction.

7. Energy costs of connection and adaptation to the grid.

Here wil work with the same deductions as in § 6, see § 4c.

8. Need for more OCGTs.

OCGTs are the best generators to compensate for rapid variations. Their thermal efficiency is about half of that of a CCGT. OCGTs are always used because of fast changes of demand. Now the variations of wind power add to the variations of demand, which requires more often production with their low efficiency and accompanying more fuel use.

We assume that with windparks of 100 MW, resp. 200 MW and 300 MW the participation of OCGTs has to be increased by resp. 3, 6 and 10%. That we are not dealing here with a negligible complication can be illustrated with a comic remark by the CEO of the Gas Union, the main natural gas supplier in the Netherlands. While he was being interviewed on Dutch TV about the huge activity of constructing new gas pipe lines, he said: “It is because all that wind takes so much gas.”

In our computations we reduced the effective η of the conventional plant according to the said percentages with the η of the OCGTs, for which we took η_OCGt = 0,32.

The other factors mentioned in §§ 4f, 4g & 4h we leave out.

9. Results & conclusions.

The result of our calculations are summarised in table 4. One must keep in mind that the conventional plant by itself is capable to fullfil the whole electricity demand. So all costs for buying the wind equipment, the costs of installation and those of the extra cables and net adaptation are extra. (See Appendix for the algorithmes.)

Table 4.
Fuel saving and CO₂ emission saving through windparks according to different models and including other relevant factors.
Results for a 500 MW production provided by a modern gas fired plant with design capacity of 500 MW together with a windpark with name plate capacity, V, near Schiphol on a normal windy day.

V	100 MW wind	200 MW wind	300 MW wind
Old model	4,2%	8,3%	12,5%
Quasi stationary ≈ current	3,5%	7,1%	10,7%
Including ‘cycling’	1,4%	2,9%	4,4%
Ibid. incl. lifetime 30 yr.	1,0%	2,0%	3,1%
Same lifetime 15 yr.	0,6%	1,2%	1,9%
Same (30 yr.) + OCGT	−0,3%	−0,5%	−1,0%
Same (15 yr.) + OCGT15	−0,8%	−1,4%	−2,3%

It is clear. The alleged savings provided by windparks that could cover 20%, 40% or 60% of the electricity demand during favourable winds are not just negligible, they are even negative, when the most relevant factors are taken into account. As we remarked before, there is substantial evidence that a life time of 15 year is not an exaggeration. We mentioned the park that had to be renewed after 12 year. That was an on shore park. The parks to be constructed off shore operate under more difficult circumstances. Therefore we conclude:

NON-SUSTAINABLE. A 300 MW nameplate windpark near Schiphol on August 28, 2011, a normal windy day, during 21,5 h would have increased the amount of natural gas needed for the electricity production of 500 MW with 47150 m3 gas. This would have caused an extra emission of 117,9 ton CO2 into the atmosphere.

The windparks do not fulfill ‘sustainable’ objectives. They cost more fuel than they save and they cause no CO₂ saving, in the contrary they increase our environmental ‘foot print’.

A decision to invest thousands of millions Euros in the construction of windparks ‘to save fossil fuel and to reduce CO₂ emission’ is irresponsible. There are no savings, THERE IS LOSS!

We do not consider it likely that more knowledge of the factors influencing the present outcomes would change our results appreciably.

Nieuwegein, October 7, 2011.

original:  www.clepair.net/windSchiphol.html
Click here to go to APPENDIX

References & notes.

• This web page article is aimed at readers with some background knowledge. It is a short version of the Dutch report Gas, Wind en CO₂ op Schiphol. De crash van de windmolens. (The Dutch report tries to explain the subject to the laymen.)

1. Dijkema, Z. Lukszo, A. Verkooijen, L. de Vries & M. Weijnen: De regelbaarheid van elektriciteitscentrales; quickscan in opdracht van het Ministerie van Economische Zaken; TU Delft, 20 April 2009.
2. Oscar Vlijmen.
3. CO₂-PROFIEL
4. K. de Groot & C. le Pair: The hidden fuel costs of wind generated electricity. Also: SPIL 263 – 264 (2009) p.15 ff.
5. C. le Pair & K. de Groot: The impact of wind generated electricity on fossil fuel consumption.
6. F. Udo, K. de Groot & C. le Pair: Wind turbines as a source of energy.
7. KEMA: priv. comm.
8. F. Udo: Wind energy in the Irish Power System.
9. Wind record Schiphol
10. Kent Hawkins: Wind Integration Realities: Case Studies of the Netherlands and of Colorado.
11. W. Post: Wind power and CO₂ emissions.
12. Hugh Sharman: Wind energy, the case of Denmark.
13. BENTEK Energy: How less became more: Wind power and unintended consequences in the Colorado energy market.
14. SEAI.
15. In discussions among us (De Groot, Udo and myself) it has been asked whether the data of ref.7 should not be interpreted as ‘1% more than the current model’? We think not. Remember the ‘car in city’ argument above. Nevertheless, we have also done the calculations using that assumption. In this case the outcomes taking into account the other factors as well are for a 15 years life time:

    savings of resp. 1,2%; 2,6% and 3,7% for 100 MW wind, 200 MW wind and 300 MW wind, i.e. 20%, 40% and 60% of the total demand capacity in the form of windmills.

    These are also absurd low savings in the view of the economics of electricity production.

16. CBS Statline.

Click here to go to original paper (which may have been updated since posting here).
Nederlandse origineel (pdf):  Gas, Wind en CO₂ op Schiphol: De crash van de windmolens