Swaminathan N., Bray K.N.C. (eds.) Turbulent Premixed Flames

Подождите немного. Документ загружается.

4.2 Application of Lean Flames in Aero Gas Turbines 333

academic researchers. Access to industrial test facilities may enable researchers to

bring their expertise to bear in a more realistic and safe environment.

Although it is not necessary to incorporate thermal management into research

fuel injectors, they should be so designed that insulating air gaps are simulated by

solid metal. This is also true for CFD studies. An additional total thickness of about

4 mm is required to allow for heat shields on both sides of a fuel passage.

Better understanding of fuel–air mixing and means to disperse fuel rapidly are

needed. The effects of pressure and temperature, although expensive to replicate,

cannot be ignored. Both have a strong impact on liquid atomisation, dispersion, and

evaporation.

There is a need for experiments to validate and support the development of a

better CFD code. Of particular importance is the description of the fuel atomisation

process. There are two difﬁculties here:

1. The process is likely to be strongly inﬂuenced by the injector conﬁguration.

There are many variables in the injector geometry, and although airﬂow pattern

is not affected by scale, fuel jets and drops do not scale in the same way.

2. We are primarily interested in the interaction of fuel and air at high power. The

very high fuel ﬂows in gas turbines produce sprays so dense that they are not

amenable to measurement with techniques and technologies currently available.

Finally there is an obstacle to collaboration arising from the commercial sensitivity

of this technology. Industry will not allow publication of research it has co-funded

that investigates geometries close to the envisaged product. To do so would threaten

competitiveness. For the same reason, industry will not divulge details of its future

designs to aid research that it is not funding. Thus, in practice, either the science-

based community must sacriﬁce its right to publish freely or industry will hold it

at arms’ length, and the effective application of precious skills to the problem of

improving gas turbines will be limited.

It may appear that the message is ‘It is all too difﬁcult, don’t attempt it!’ That is

not the view of this author, but 20 years spent in managing collaborative research of

this kind has yielded some lessons. The message is ‘Proceed with care as the problem

is complex and challenging, yet crucial to understand!’

CFD. One of the patriarchs of gas turbine combustion, Professor A.H. Lefebvre,

would assist his students to grasp the place of CFD in this ﬁeld by bringing into his

lecture room a T shirt, on the back of which, he told them, had been printed the

major contributions to gas turbine combustion made by CFD. He would then hold

it up to the audience. It was completely white!

Since the turn of the century that situation has changed. The massive comput-

ing power of parallel processors combined with the patient efforts of many highly

competent people over the past 40 years are at last bearing fruit. The aerodynamic

ﬂow ﬁeld generated by complex geometries such as lean-burn fuel injectors can now

be calculated reliably. A great deal of work has been done to improve both combus-

tion models and chemistry–turbulence interaction models. LES provides a greatly

improved model of unsteadiness and turbulence, but requires careful meshing, and

the cost and time expended in its application is a serious problem for industry. Sci-

entiﬁcally, the subgrid modelling is challenging. At present it is common practice to

334 Lean Flames in Practice

use both LES and URANS to model unsteady phenomena. Research is ongoing in

many universities to improve these models.

The importance of appropriate boundary conditions cannot be overemphasised.

As it is comparatively easy to measure the aerodynamic ﬂow ﬁeld external to the

combustor, but very difﬁcult inside, and as CFD offers greatest beneﬁt when applied

to the complex, internal ﬂow ﬁeld, it is good practice to determine the boundary

conditions experimentally as close to the ﬂame space as possible. The most serious

deﬁciency of CFD is in the deﬁnition of spray boundary conditions and modelling,

and, as mentioned in the previous section, this is also very difﬁcult experimentally.

The empirical correlation [216] now used to estimate the SMD was generated from

fuel injectors far different from those now under development for emissions control.

Advanced, two-phase ﬂow methods to predict primary break-up of the fuel spray

are under development, but it will be some time before we can conﬁdently apply

these techniques.

The way we use CFD in combustion is to anchor it to existing experimental

data and then use it for predictions cautiously. Availability of experimental data is

extremely important as it enables the selection of the most appropriate model (there

are about 10 turbulence models to choose from!)

Good work has been done in the representation of evaporating two-phase ﬂow

and fuel-spray development in cross-stream injectors [214, 217]. Another paper [218]

is concerned with the break-up of a diesel injector spray and provides a good ex-

ample of advanced two-phase ﬂow modelling. The sprays under consideration are

extremely dense, so experimental validation is not possible. Thus, aided by exper-

imental validation, CFD is ready to be applied to the design of premixing ducts.

Although not completely reliable, it is far better than guessing.

We are not able to predict CO emissions reliably. The empirical design rules

for conventional combustors are adequate, but with lean burn, CO formation and

consumption is a problem. Several effects contribute to this difﬁculty (spray bound-

ary conditions, ﬁnite-rate chemistry, complexity of the partially and fully premixed

combustion modelling, etc). Premixed combustion modelling is more difﬁcult than

diffusion-ﬂame modelling, which makes application of CFD to lean burn more of

a challenge. Some recent progress on this front is discussed in Chapter 2 of this

book.

Interaction between pilot and main zones is important for both emissions control

and ﬂame stability. It is likely that unsteady ﬂow effects are important here, and it

may be asking too much of CFD to handle the problem. However the parallel use of

CFD with experiment can be of enormous help in interpreting experimental results

and guiding the aerothermal designer in deﬁning modiﬁcations that will improve

performance. The usefulness of CFD depends on the reliability with which the ﬂow

ﬁeld and fuel distribution, among many other factors, can be modelled.

One of the advantages of CFD over experiment is that it is not limited by

operating conditions: The problem is that neither can it be validated at high pressure

and temperature. It can only be shown to work by experience in application. Here

we must be wary of our own selectivity. When CFD yields a result contrary to

observation of engine behaviour, there is a tendency either to adjust something that

cannot be reliably estimated, for example SMD, select a different turbulence model

to achieve agreement, or simply ignore and move on. When CFD shows a trend that

4.3 Application of Lean Flames in Stationary Gas Turbines 335

matches engine behaviour, the results are naturally trumpeted from the rooftops.

We all love to achieve something, and publishing our achievements is the only way

of ﬁnancing further work. But this normal behaviour is biased. The bias may be

unavoidable, but let us at least not deceive ourselves. There is no such thing as a bad

result in research, if competently conducted. We learn as much from our failures as

our successes – but only if we give our failures equal weight.

CFD provides a useful approximation. We are often unable to estimate the level

of conﬁdence in our CFD predictions, and there is therefore a natural reluctance

to rely on CFD to support initial design and component development. It is in these

applications where the potential cost and time savings are greatest, and there is a

strong expectation that continued improvement of the models together with experi-

ence in their application will yield signiﬁcant beneﬁts. Finally, with reference to time

scale for improvement, for the past 30 years, the breakthrough that would transform

the situation has always been from 2 to 5 years away. This has something to do with

funding horizons and the patience of stakeholders. It is encountered in every ﬁeld of

research. The appropriate response is a wry smile!

4.3 Application of Lean Flames in Stationary Gas Turbines

By B. Jones

Stationary gas turbines, including those used on board ships as auxiliary power

generators or main propulsion units, vary greatly in both size and design. Simple

cycle machines range in power from 100 kW to a little over 300 MW, and combined

cycle machines produce up to 450 MW.

The absence of airworthiness constraints described in the introductory subsec-

tion of Section 4.2 opens wider opportunities for innovation in the design of com-

bustion systems for stationary engines. Land-based machines are subject to stringent

emissions regulation because they often operate continuously at full load close to

populated areas. The aeroengine industry has struggled for the past 35 years to make

signiﬁcant reductions in NO

emissions, the advances in combustion technology be-

ing largely negated by increases in pressure ratio and turbine inlet temperature. In

the industrial sector NO

has been reduced by 90% to 95% by the application of

lean combustion.

By integrating the gas turbine into a combined cycle in which exhaust heat

is used to produce additional power from a steam turbine, cycle efﬁciency can be

increased from that of an aeroengine – around 40% – to almost 60%. The efﬁcient

use of heat that would otherwise be wasted inﬂuences the choice of gas generator

cycle. There is little to be gained from increased pressure ratio or very high TET,

both of which greatly increase cost. Pressure ratios of stationary gas turbines are

thus usually around 15:1, although there are engines that operate at more than

double this pressure. Although aeroderivative engine combustors can be designed

without regard to airworthiness, they have to be compatible with the mechanical

design of the original engine core. Efforts to radically reduce emissions have, in the

past two decades, led to a multiplicity of new designs that bear little resemblance

to combustors used towards the close of the 20th century. The constraints of size,

weight, and ﬂight safety that so constrain the design of aeroengine combustors has

led to convergence between competing designs to something approaching a single,

336 Lean Flames in Practice

Figure 4.58. R-R industrial RB211 two-stage gas-burning LPP combustor. (Courtesy of

Rolls-Royce plc.)

optimum solution. The absence of these constraints has led to remarkable diversity

of designs in stationary engines.

4.3.1 Common Combustor Conﬁgurations

1. A ring of tubular combustors that may be arranged within an annular casing or

in separate pressure casings mounted radially around the engine (Figs. 4.58 and

4.59). Where the tubular combustors are arranged axially, they are often angled

to allow the fuel injectors to be outboard of and upstream of the compressor

outlet, to reduce shaft length. Both conﬁgurations create serious aerodynamic

problems in relation to the diffusion and distribution of the compressor outlet

ﬂow.

2. A single annular combustor fed by one or more rings of fuel injectors; see

Fig. 4.60.

3. A single tubular combustor may be mounted radially on the side of the pressure

casing in small engines or situated outside the engine gas path in a silo adjacent

to the engine in large engines. This larger combustor may be fed by many fuel

injectors, as shown in Fig. 4.61.

4. Interstage reheat is employed in Alstom’s GT24 and GT26 engines, shown in

Fig. 4.62. The incorporation of an interstage combustor to reheat the turbine gas

increases the speciﬁc power output of the machine without raising the maximum

cycle temperature. Alstom developed a premixing fuel injection system for this

combustor that actually reduces the NO

from the primary combustor [219].

4.3 Application of Lean Flames in Stationary Gas Turbines 337

Figure 4.59. KHI M7A-03 8-MW gas turbine with tubular dry low-NO

(DLN) combustion

system. (Courtesy of Kawasaki Heavy Industries, Nishi Akashi, Japan)

The choice of conﬁguration is driven by the engine operating conditions, me-

chanical design considerations, cost, and probably by the experience of the design

team. In gas turbine design, innovation brings with it a period of trouble and high

cost. The ﬁnancial return often comes many years later. If a company has a successful

design they will ﬁrst consider scaling that. If the engine is of a high pressure ratio,

the combustor casing will be highly stressed and the conﬁguration will be selected

with that in mind. This will also place a heavy demand on cooling, so the combustor

wall area will need to be minimised. A silo or annular combustor will be preferred.

A company with limited test facilities will want to avoid the use of an annular com-

bustor if possible because the cost of rig testing is enormous compared with testing

one of a set of tubular combustors.

Figure 4.60. Alstom annular DLN combustor showing multiple premixed fuel injectors.

(Courtesy of Alstom Power)

338 Lean Flames in Practice

Figure 4.61. Large engine silo combustor with multiple fuel injectors.

4.3.2 Fuels

Gas turbines are less tolerant to fuel variations than either diesel engines or boilers

for the following reasons. Turbine blades cannot tolerate contaminants like sodium,

vanadium, potassium, or ash, either because of high-temperature corrosion or block-

age of small cooling holes by solid deposits. Nor can they tolerate large changes in

the caloriﬁc value of the fuel supplied. Low-caloriﬁc-value gas from coke ovens and

Figure 4.62. Lean-burn reheat combustion system. (Courtesy of Alstom Power.)

4.3 Application of Lean Flames in Stationary Gas Turbines 339

blast furnaces [coke-oven gas (COG), blast furnace gas (BFG), and Linz–Donawitz

gas (LDG)] has a low burning velocity, approximately proportional to its heating

value by volume, which may be less than 10% that of natural gas. Syngas has a much

lower heating value than natural gas but a high burning velocity, as it can contain

more than 40% hydrogen. These low-caloriﬁc-value gases require the use of special

injectors to accommodate an increased volume ﬂow rate. The dilution of the fuel

by nitrogen and CO

reduces reaction rate, and combustion efﬁciency will be re-

duced at low power. Due consideration needs to be given to the power needed to

raise the pressure of the greater mass of fuel and the size of the injectors required

to accommodate it. The fuel supply pressure must be around 15 bars higher than

the combustor pressure to accommodate the losses across the fuel control system

and fuel injectors. Low-caloriﬁc-value gas is often mixed with natural gas to contain

these problems. It will be apparent that there is a limit to the application of lean

ﬂames with these fuels. Lean combustion is already close to weak extinction and

cannot tolerate a large reduction in ﬂame speed or ﬂame temperature. LPP systems

cannot tolerate a signiﬁcant increase in ﬂame speed associated with the presence of

hydrogen because of ﬂashback.

Most engines used for power generation burn natural gas, and a signiﬁcant pro-

portion of these engines have the ability to burn both gas and liquid fuels. Sometimes

there is a requirement to operate with gas mixtures, including propane, liqueﬁed

natural gas (LNG), or syngas to take advantage of low-cost alternative fuels. Gas

turbines operating on gaseous fuels cannot tolerate liquid HCs in the fuel, and hence

fuel preheating is commonly employed in cold climates or where gas contains heav-

ier fractions. There is also widespread demand for the ability to burn light fuel oil as

a back-up should the gas supply fail. Diesel fuel is the only fuel available on board

ship.

Lean-burn combustions systems are more susceptible to changes in gas-fuel

properties. Whereas a diffusion burning combustor can typically tolerate a ±20%

variation in Wobbe index, LPP combustors can accept only a ±10% variation of this

index. The Wobbe index, commonly used to denote changes in gas-fuel properties,

is deﬁned as the ratio of higher heating value in millijoules per normal cubic meter

(Nm

, i.e., at standard temperature and pressure) to the square root of the speciﬁc

gravity of the fuel gas. The signiﬁcance of the Wobbe index is that, for a given power

output and injector effective ﬂow area, a fuel with a high Wobbe index will require

a lower injector pressure drop, increasing the likelihood of combustion instability.

Depending on the injector design, it may also inﬂuence the degree of premixing

achieved. The Wobbe index for natural gas varies with the source but is typically

around 50 (1250 in BTU/ft

used in the United States), whereas propane has a

Wobbe index of 80 (2000 in BTU/ft

). Auto-ignition properties of fuels also need to

be considered and are discussed in Subsection 4.3.9.

4.3.3 Water Injection

The simplest way to control NO

is to reduce maximum ﬂame temperature by

injecting either water or steam. The ratio of water to fuel is typically around 1, and

this ratio can go up to 1.6 with steam. There is an increase in power of up to 5%

because more fuel can be burned without exceeding the HP turbine temperature

340 Lean Flames in Practice

limit, increasing the enthalpy, density, and speciﬁc heat of the ﬂow through the

turbine. The disadvantages, which drive the continued research of dry low-NO

(DLN) solutions, are as follows:

1. The highly puriﬁed water required is not always available in sufﬁcient quantities

and it is costly. A 100-MW engine might consume more than 25 tons/h. There

are places like California where there is high demand for power, there is little

water, and there are stringent emissions regulations necessitated by the climate

(see Subsection 4.3.4). Cycle efﬁciency is reduced by up to 4%.

2. CO emission increases, especially at part load, and may exceed local air-quality

requirements.

3. Water and steam injection reduce component life and increase maintenance

costs. Prandtl number and Reynolds number increase, thus increasing the heat

transfer coefﬁcient in the combustion gas. The injection of water or steam prob-

ably also modiﬁes the combustor outlet temperature proﬁle.

4. Continuous monitoring of exhaust emissions is required, whereas for DLN

systems only an annual veriﬁcation test is required. However, the DLN systems

require continuous monitoring of combustion instability.

Offset against the costs associated with water or steam injection is that it avoids

the problems that commonly occur in lean-burn systems, especially when combus-

tion of diesel fuel is required. These are auto-ignition and ﬂashback, combustion

instability, and fuel system complexity. Maintenance costs and engine downtime are

also higher with engines ﬁtted with LPP DLN combustors. According to [220, p. 426],

maintenance costs for engines using DLN combustors are 40% higher than for en-

gines using diffusion burning. What is probably more serious in some installations

are unscheduled shutdowns, usually caused by the onset of combustion instability.

This problem is receiving a lot of attention and has been with us for over 20 years.

The solution may not be imminent; however, we trust that it will be solved in due

course. The subject of combustion instability is treated in depth in Chapter 3 of this

book.

4.3.4 Emissions Regulations

There is no single document that deﬁnes regulations for stationary gas turbines.

This subject is complex, and only those who market gas turbines or act as advisors

to the industry can justify the time required to understand it and keep up with its

development. The author of this section is indebted to the Industrial Power Division

of Rolls-Royce plc in Montreal for providing advice and the references needed

for this text. The documents are complex, and the author’s research has not been

exhaustive. The aim of this section is only to explain the targets to be met by the

combustion engineer.

The EPA website, www.epa/nscep/, provides a free service, allowing any docu-

ment published by the EPA to be downloaded as pdf ﬁles. This source is so abundant

that ﬁnding something can be difﬁcult, even with the help of the advanced search

facility on the site.

The minimum U.S. federal standard for NO

and SO

control is deﬁned in

[221], which also explains the rationale behind the legislation in great detail. The

4.3 Application of Lean Flames in Stationary Gas Turbines 341

emissions controlled are NO

and SO

, which come from sulphur in the fuel. CO

and particulate emissions are regulated separately but not across the entire U.S.

continent. The federal regulations require gas turbines of greater than 110 MW to

emit less than 15 ppmv NO

and gas turbines of between 25 and 110 MW to emit less

than 25 ppmv. Regulations vary from one country to another, state to state in the

United States, and in California even district to district. California and Florida apply

the most stringent regulations (see, for example, [222, 223]). The combination of

geographic and climatic conditions in California results in the production of highly

toxic ozone at ground level by the action of UV radiation on NO

. In Florida it is

because of the large orange-growing industry. Small concentrations of NO

discolour

the skins of oranges. Permitted concentrations of NO

depend also on power plant

efﬁciency and size. Local regulations are updated in response to the opening of new

plant of a type not previously covered by legislation in the region and in response

to the introduction of new technology. The use of best available control technology

(BACT) and best available retroﬁt control technology (BARCT) is demanded by

some U.S. states. This drives competition and of course requires amendment of

legislation, making it difﬁcult to keep up to date. For example, [223] was amended

seven times between 1994, when it was ﬁrst issued, and 2009. Failure to match BACT

can exclude a company from certain sectors of the market. Exceptions are permitted

to allow for off-design performance during start-up, which for some types of plant

can take 3 to 4 hours, and for malfunction, but these exceptions must be recorded

and reported. Sometimes heavy ﬁnes are incurred.

The requirements for the Sacramento metropolitan area [222] are summarised

in the following list, as an example, and NO

limits are corrected to 15% O

,dry

basis:

1. Gaseous fuel ﬁring

42 ppmv for units ≥ 0.3 MW and < 2.9 MW

42 ppmv for units ≥ 2.9 MW and operated < 877 h/year

25 ppmv for units ≥ 2.9 MW and < 10 MW and operated ≥ 877 h/year

15 ppmv for units ≥ 10 MW, operated ≥ 877 h/year, without SCR

9 ppmv for units ≥ 10 MW, operated ≥ 877 h/year, with SCR

2. Liquid Fuel Firing

65 ppmv for units ≥ 0.3 MW and < 10 MW

65 ppmv for units ≥ 10 MW and operated < 877 h/year

42 ppmv for units ≥ 10 MW, operated ≥ 877 h/year, without SCR

25 ppmv for units ≥ 10 MW, operated ≥ 877 h/year, with SCR

In [222] these limits are compared with the limits for other districts in Califor-

nia. They are typical, but the regulations for the San Joaquin Valley are the most

stringent, requiring 5-ppmv NO

for a new plant using certain gas turbine types. This

would usually require an exhaust clean-up of some kind. Regulations are extremely

detailed, sometimes specifying different limits for speciﬁc engine models. The San

Joaquin Valley Authority imposes ﬁnes for non-compliance of up to U.S$75 000 per

ton of NO

, increasing to U.S.$100 000 per ton [223] in 2014.

CO regulations also vary from place to place; 50 ppmv is the most common value

applied, but 25 ppmv is required in the San Joaquin Valley for some installations.

342 Lean Flames in Practice

The regulatory limits applicable to land-based gas turbines within the EU are

speciﬁed in [224]. The rules apply to plants of greater than 50-MW heat input.

for a gas-ﬁred plant must be below 24 ppmv (15% O

), 53 ppmv for liquid

fuel. As in the United States, local air-quality rules may be more stringent than the

EU regulations. Local air-quality rules often take into account the number of hours

operated annually. There are complex variations related to the age of the power plant

that require study. Emissions regulations sometimes specify pollution in mg/Nm

50-mg/Nm

is equivalent to 24.3 ppmv corrected to 15% O

. CO is not yet

controlled by the EC, but is expected to be included in the near future.

Another variable is the lower limit of the power range over which emission

regulations are applied. This may be anywhere from start-up to 70% power, with

varying degrees of alleviation for warm-up and variation in ambient temperature.

The rules are different for different types of plants because of the large differences

in warm-up time and use.

4.3.5 Available NO

Control Technologies

Lean DI, unaided by steam or water, cannot achieve the required NO

and CO

control. There is too much excess air in the exhaust of a gas turbine for a three-way

catalyst to work, and the exhaust temperatures from gas turbines preclude the use

of most alternative options of employing catalysts. Selective catalytic reduction of

using ammonia injected into the exhaust upstream of a catalyst can achieve NO

below 5 ppm. Ammonia is, however, a hazardous substance, and its accidental release

in a populated area would have serious consequences. Lean premixed combustion

is the most widely used and most practical remedy. Chapter 7 of [223] provides a

thorough assessment of chemical means of exhaust clean-up for a gas turbine plant.

I. Gas-burning systems:LPP combustors are now offered as standard equipments for

most stationary gas turbine engines using gaseous fuels.

1. Heavyweight engines

Asea–Brown–Boveri (ABB) demonstrated their ﬁrst LPP combustor in a GT13

150-MW service engine in 1984. The combustor was of the silo type with multiple

fuel injectors, and six engines entered service. NO

below 100 ppm was achieved

with low CO, a commendable achievement at that time [219]. In 1993 Alstom,

which had bought ABB, introduced the EV burner, which is a conical premixing

burner, into service [219], which proved so successful that it is now used in all

Alstom engines (Fig. 4.63). The burner is able to burn fuels ranging from syngas,

which has a high hydrogen content yielding a high ﬂame speed and susceptiblity

to ﬂashback, to number two diesel oil. NO

below 25 ppmv (corrected to 15%

) at full power and CO of 4 ppmv is claimed.

Alstom introduced a variant of the EV burner, the SEV, for use in their intertur-

bine combustors for the GT24 and 26 engines [219]. This fuel–air mixing device

is modiﬁed to avoid auto-ignition at the high inlet temperature to which it is

exposed in this application. The fuel is admitted with compressor delivery air to

achieve a degree of premixing at lower temperature and to increase the thermal

mass to delay fuel heating to its auto-ignition temperature. The injector also