Vous êtes sur la page 1sur 18

Replacing Black Start Diesel Generators with Energy Storage Systems

The grid has to be prepared for the extreme case of a complete blackout. This, however,
is complicated by the fact that most power plants require a lot of time and energy to prepare
for operation: for example it can take up to an entire day to get a big coal fired plant to
maximal output from a cold state, and it can not be done without considerable energy
investment. It is crucial then to have power plants that are able to start quickly and
independently of grid power. The ability of independent startup is called black start capability,
and because of the extra investment it requires, usually only a few plants have this capability
in a national grid. Black start plants are traditionally gas turbine plants, since they can be
started in about 15-20 minutes, or even faster in emergencies. The way a gas turbine plant
normally starts is that the attached generator or separate starter motor is energized from the
grid, it cranks up the turbine that then gradually starts to supply power as the pressure and
the heat builds up. Without grid power, the required electricity comes from a standalone
diesel generator set, which is in turn started from a small battery - similarly as a big truck
engine. By regulation, the generator set must have enough fuel for at least three black start
1, and the plant itself for a few days of continuous operation.
Normally these fast ramping plants are only used a few hours each year, in times of
other plants outages or extreme high loads, since they are a lot less economical than base
load plants. Even in these rare times, the plant is started using grid power, the black start
diesel generator (BSDG) is often never used in its entire lifetime, apart from drills and
maintenance runs. It could be economical then to replace it with a battery pack with the
necessary performance and capacity to supply the three black start attempts, and otherwise
use it to provide other auxiliary services to the grid.
This paper assesses the feasibility of different battery energy storage systems that would
provide black start capability as well as frequency containment reserve. A model is built to
find the ideal capacity and the costs associated with providing the different services.
The black start process
Black start is the process of bootstrapping a power plant into operation without relying on
outside electricity, with the intention of restoring grid power after a system wide blackout.
The smartest way to black start the entire grid is to fire up a few fast-ramping power plants
first, use their power to start other fast-ramping plants without black start capability, start
slower-ramping base load plants and gradually restore service. Very often hydroelectric
plants would provide this capability, since they can be put into operation very quickly and
without major energy investment. In countries without adequate hydropower, such as
Hungary, this is carried out by gas turbine power plants with black start diesel generators.
Usually the process is as follows:
1. A small battery starts a BSDG located in a gas turbine plant
2. The BDSGs power starts the gas turbine
3. A transmission line towards other non-capable plants is energized
4. A base plant is brought into operation using the grid power
5. The base plants power is then used to restart all other power plants in the system
6. Power is reapplied to the distribution network and sent to the consumers - gradually

http://www.hadmernok.hu/142_01_berekl.pdf

Page 1
! of !18

It is important not to restore service everywhere at the same time, because the sudden
huge load could trip the power grid again. In larger grids, the procedure usually involves
reapplying service to multiple independent islands first, and then synchronizing them.
There are three gas turbine power plants in Hungary that provide reserve capacity and
are also key in black starting the national grid: Dunamenti, Gnyi, and Lrinci power plants.
Black Start Diesel Generators
The diesel generator sets used as black start generators tend to be in the 2.5 to 5 MW
range2, depending on the type and quantity of gas turbines it has to be able to start. These
units, usually called gensets for short, are generally compact systems built into a shipping
container, and are used for a wide variety of applications, as backup generators for
hospitals, data centers, mobile prime power sources for mines, small islands, etc. One
container would include the diesel engine (usually a large displacement 8 or 16 cylinder unit
in this performance range3), a transmission, the generator, control units, and fuel
management system. Turnkey solutions are available from multiple manufacturers, and it
can be considered a mature, reliable technology. Such as: MTU 0080-4000 DS, GE 16V250.
Energy Storage Systems
Technically there seems to be no reason why BSDGs could not be replaced with battery
packs - even though there is little precedent of such installations in Europe.4 The goal of this
paper is to assess the economic feasibility of such a solution with respect to all the
alternative applications the battery based solution would allow. The energy storage system
(ESS) would have to have enough charge left at all times to provide the necessary power
continuously for the duration of at least three black start attempts. It may be reasonable to
build extra capacity to obtain the capability of providing ancillary services to the grid, such as
peak shaving, fast regulation, etc. or defer upgrading transmission and distribution (T&D)
equipment.
The minimum criteria and the operational environment of the system to be designed:
Highly reliable - at least as much as the diesel genset.
Has the required output power to bootstrap the gas turbine.
Has the necessary capacity to power three start attempts any time.
No vibrations, protection from punctures and other mechanical damage.
Well defined and narrow operational temperature range.
Virtually unlimited space - dimensions and weight are irrelevant.
Easy to reach location, availability of skilled personnel: regular maintenance and
monitoring is possible.
Since there is a minimum charge level required for black start capability, the battery
would never be deep cycled, apart from drills and actual crises.
For the same reason, roundtrip efficiency is also irrelevant. (Not for secondary
purposes though!)
Low self discharge would be a relevant advantage.
2

http://www.ryanwilks.com.au/wp-content/uploads/2010/02/Colongra.pdf

http://www.mtuonsiteenergy.com/products/diesel-generator-sets/mtu-4000-ds/

http://www.energystorageexchange.org/projects

Page 2
! of !18

To sum up, the requirement is a high capacity battery for the lowest overall cost. Since
the discharge rate would not exceed 1C, the power requirement is not restricting the
chemistry, almost all solutions are capable of sustaining this rate.
Types of ESS
The cheapest of all systems is pumped energy storage: when energy is stored in water
reservoirs or as high pressure air pumped into caves - these however require special
environmental conditions. I will focus solely on solutions that can be deployed anywhere and
can be set up in containers on the power plant or substation premises.
The most conventional and wide spread battery technology is lead acid. Lead acid
batteries are relatively cheap, readily available, and there are many precedents of grid-scale
installations. The common drawbacks of the lead acid technology is that they require some
maintenance due to the escaping of hydrogen gas when the battery is overcharged, they
only work ideally in one set orientation, they are prone to spilling, and they degrade quickly if
deep cycled. They are ideal as motor vehicle starter batteries, because in such an
application they are rarely discharged below 80% SOC, and their large surge power
capabilities are needed. The nature of the discussed application eliminates most of the
drawbacks, therefore this chemistry is a good candidate for black start battery. When lead
acid is used in frequency regulation for its high ramping capability, the storage system is
often limited to ~20% DOD to maximize battery lifetime. Since high temperatures can
drastically reduce lifetime, it is important to provide adequate thermal management in the
storage containers. Lead acid batteries operated at partial state of charge also require
refresh cycles to dissolve sulphate crystals that have accumulated on the negative electrode
and replenish the capacity of the battery. This would cause a downtime of a few hours every
~10 days in grid support applications.5
Another common and increasingly popular type is lithium-ion. In high power
applications, Li-ion is competitive when its advantages such as tiny memory effect, low
weight, and low self discharge are relevant. Therefore it is widely used with solar/wind
installations, for frequency regulation purposes, for peak shaving, and for back up power
where outages are frequent. The main drawback of the lithium based technologies are their
high price. It is important to note that lithium-ion is probably the most researched technology,
and its production costs and capabilities are improving at a high pace, especially with
economies of scale kicking in. There is already precedent for using Li-ion batteries for black
start in Eisenhttenstadt, Germany: a 2.8 MW, 720 kWh installation that can bootstrap a 153
MW gas turbine into operation. Li-ion batteries are better suited for deep cycling, and in
general longer lasting than lead acid based solutions.
Nickel-Cadmium, a once popular chemistry for rechargeable AA batteries now banned
by the European Union is powering one of the worlds biggest batteries in Fairbanks, Alaska.
The installation demonstrates reliability and has a power rating of 40MW, which it can
sustain for about 7 minutes - the giant UPS is used as a backup for the too frequently failing
grid. A big advantage of Ni-Cd batteries is that they tolerate both high discharge rates (up to
50C even, although ideal is below 15C) and deep cycling, making it ideal for short term
backup solutions, even if outages occur fairly often. It is however substantially more
expensive than lead acid solutions of equal capacity, and its benefits are irrelevant, as in the
black start application the discharge rate would be around 1C, while deep discharge would
be extremely rare.

http://www.ultrabattery.com/technology/ultrabattery-performance-benefits/

Page 3
! of !18

Nickel-Metal Hydride is the technology that replaced Ni-Ca in the rechargeable AA


battery market, as it is more eco friendly and has higher energy density. It is also a popular
chemistry for producing larger battery packs for electric or hybrid vehicles, being the chosen
technology of Toyota and Honda. Nowadays most purely electric vehicles use Li-ion despite
the safety concerns and higher price, because it has even higher specific energy. Ni-MH was
never popular for large stationary batteries due to low cycle life and irrelevant advantages - it
is mentioned in this paper for the sake of completeness.
Turning away from chemical batteries, compressed air energy storage (CAES) is an
inexpensive solution due to cheaply available tanks and compressors, whose main issue is
low roundtrip efficiency. This however can be overcome with the latest technologies that
sprays water into the cylinder in the compressor and stores the resulting hot water. The
energy density of such a storage is considerably lower, looking at LightSail Energys solution
for example, it stores 750 kWh in 45 foot shipping containers. The storage units themselves
are however quite simple carbon fibre reinforced high pressure air tanks, cheaper and longer
lasting compared to batteries. As of power output, each module is capable of 500 kW,
therefore 5 units would be needed to supply the turbine startup procedure. These units
supply 3 phase AC power that could be directly applicable to the gas turbines starter motor.
Advantages of this type of EES is the virtually zero self-discharge, long service life (20+
years for the compressors, even more for the storage tanks), and lack of hazardous
materials. Disadvantage is the lack of commercial availability as of yet - the few existing
CAES plants are in the tens or hundreds of MWh range, and use underground caverns or
caves versus LightSails aboveground portable solution. Many of these plants would use the
stored compressed air to support power generation via natural gas turbines instead of simply
turning a generator trough an expander. This sort of use however is only economical with
larger amounts of compressed air stored in some natural cavity, therefore it does not apply
to the case under investigation.
Cryogenic energy storage (CES) is another novel idea on the market, currently under
testing in the United Kingdom. It uses very low temperature liquids to store energy - usually
liquid air or nitrogen at temperatures around -200 C. They use the Claude cycle to cool air
from the atmosphere to the point it liquifies, and then store it in a well insulated tank - on
atmospheric pressure. The advantages are the long service life (30+ years), low capital cost,
lack of hazardous materials, good scalability, and the major drawback is the low roundtrip
efficiency - unless some low grade heat source can be utilized to boil the liquid air. While this
solution is not yet available for purchase on the market, it may, in the future, be used for
black start energy storage capacity, because it can store large amounts of energy and has
low capital cost. CES is aimed at long term energy storage.
A readily available yet equally exciting solution is that of flow batteries. In flow batteries,
the energy is stored in the electrolytes, which are in turn stored in separate containers and
are never mixed, but brought into contact in the fuel cell that can reversibly convert chemical
to electrical energy. In these batteries the power is a function of the cell number and size,
whereas the capacity is proportional to electrolyte volume - therefore these can be handled
separately, yielding huge flexibility for customization. The lifetime of flow batteries can be
very long, because there are no solid-to-solid phase transitions. Self discharge is
nonexistent, because when the battery is not in operation, the fluids are not in contact. Flow
batteries also do not need equalization, a common problem in multi-cell battery installations.
The drawbacks of the technology are largely irrelevant when one considers installation in a
power plant: low energy density, need for pumps, control units, impossibility of scaling down
to handheld. Flow batteries are normally used for large stationary applications, and are
considered for electric vehicles since they can be quickly recharged by swapping the spent
electrolyte tanks for energized ones. The main drawback when considering it for a black
start application is that the fuel cells are expensive, therefore it is only economical for
applications where lower discharge rate is sufficient. In the given case, this could be
achieved by finding secondary applications that would require higher capacity but not higher

Page 4
! of !18

power. It would also incur serious additional capital costs, as it would require at least 400
7kW fuel cells, over 500 kg and 1 m3 each, plus the storage tanks - it would have to be
housed in a warehouse, not in a simple container array as a diesel generator set or a more
conventional battery pack.
UltraBatteries are a spinoff from valve regulated lead acid batteries, halfway between
conventional lead acid batteries and ultracapacitors. In practice this hybrid design means
replacing half of the negative electrode (traditionally lead) with carbon, which would act as
the electrode of the capacitor. The positive electrode (lead oxide) is common, and exactly
the same as in a conventional lead acid battery.
The advantages of UltraBatteries as compared to LA are longer lifetime, higher energy
efficiency, and superior charge acceptance under partial SOC conditions, while production
costs are comparable to conventional lead acid. In case of the black start capability plus
auxiliary service provider application, these properties would be increasingly important the
more the battery is oversized relative to the black start capacity requirement.
UltraBatteries are used for frequency regulation in Lyon Station, Pennsylvania, and for
renewable-based microgrid stabilization on King Island, Australia.

Ecoult recommends 60% DOD for regular cycling and 90% DOD for emergencies. They are
most efficient and longest lasting if they are used in a partial state of charge, and also
require refresh cycles less often.6
Analyzing the demand - gas turbine startup
The main objective of the system to be designed is to support the gas turbines starter
motors power needs. I analyzed the measurements of a start sequence from the 173 MW
gas turbine power plant in Lrinci, Northern Hungary. The data includes the most important
voltages, currents, powers, and turbine RPM for a two hour period with one second
resolution. The period includes a cold start of the gas turbine using the black start diesel
generator and its subsequent shutdown. The data was exported using the SPPA-T3000
control system of the power plant.
Since the BSDGs bus showed some activity even before the gensets breaker switched
on, I subtracted these initial values from all data points of genset current, power, and

http://www.ultrabattery.com/technology/ultrabattery-performance-benefits/

Page 5
! of !18

reactive power. The initial value may be explained by the auxiliary electricity consumed at
the plant. (16kW - 60 kvar, 1.45 A)
Runtime of diesel genset
1,20

GT reached nominal RPM


1,00

0,80

0,60

Syncing GT and BSDG

0,40

Flame in

0,20

BSDG started

BSDG is online

0,00

GT started
-0,20

Turbine n(PU)

Diesel P(PU)

Diesel Q(PU)

Switching off diesel

The key takeaways from analyzing the startup demand is that the power output of the
diesel genset peaks at just below 2 MW, and that the total consumption of one startup
attempt is around 170 kWh. Therefore the replacing ESS must have at least thrice that as
capacity and the same or higher as rated power. It is essential to also have a safety margin
on capacity in case of extended standby and also not to completely deplete the battery
system, not to mention taking into account that the capacity will somewhat decrease over
the systems lifespan due to aging. According to the data, there is a 16 kW base load, that is
most likely the standby consumption of the plant. In order to provide a 24 hour UPS service
as well as cranking power for the gas turbine, that is additional 384 kWh.
Drafting the solution - black start only
If we ignore the potentially profitable auxiliary services, the ESS has to have a rated
power of 2.5 MW and a capacity of 1 MWh. (Three times 170 kWh + 385 kWh with a 10%
safety margin.) The major costs of building such a system are the following:
1. Battery
2. Inverter
3. Control system
4. Racking, containers for housing
5. Cables and infrastructure
The price of the inverter and the control system is the same for all battery chemistries,
around $0.25/W7 altogether in such industrial applications. There are integrated solutions

http://inpressco.com/wp-content/uploads/2014/06/Paper1281859-1861.pdf
https://www.civicsolar.com/support/installer/articles/balance-system-bos-solar-pv
7

Page 6
! of !18

available in units of 500 kW or even 2.5 MW8. These would be turnkey solutions built into a
smaller portable shipping container.
The cost of racking, housing, cables, and infrastructure depend on how compact the
battery is, that is determined by the energy density of the different chemistries. (Energy
density is energy per unit volume: a Wh/m3 value.) For a lead-acid battery, it is ~0.34 MJ/liter
(based on a car battery), which means, for the planned 1 MWh = 3600 MJ, 3600/0.34 = 10.6
m3 is needed. The same for Li-ion is ~1.8 MJ/liter (based on the Panasonic 18650 cells), so
3600/1.6 = 2 m3 would suffice. By comparison, the internal volume of a 20 foot ISO container
is 33.1 m3, that of a 40 foot one is 67.5 m3.
Of course when packing so many cells and modules, and volume in not expensive, it
makes sense to pack them less tightly so that excess heat is easier to remove and individual
modules are accessible for replacement. The takeaway from the rudimentary calculations
above is that however much space in needed for lead-acid, Li-ion will likely suffice with one
fifth.
Looking at some market solutions, Mitsubishis 2 MW, 0.8 MWh integrated Li-ion based
ESS fits into three 40 foot containers for example, weighting ~22 tons each.9 (The maximum
gross weight of an ISO container of any size is limited to 30.4 tons, which also limits tighter
packing.) Therefore 1.6 m3 of Li-ion batteries are packed into ~169 m3, assuming that the
inverter takes half of a container.

Solar Power Smoothing and Energy Shifting w/ Ecoults 20


foot, 500 MWh UltraBatteries in New Mexico, USA
Ecoults UltraBattery Storage Blocks use 20 foot containers for a capacity of 0.5 MWh
and rated power of 250 kW each. These units do not include inverters, but do have
controllers and integrated ventilation/temperature control. Therefore the required capacity
would take up 2 containers, and the power electronics a third one - similarly to the Li-ion
design. (However this design would not provide sufficiently high output power.) In this case,
10.6 m3 of batteries are packed into 66 m3.10

http://www.solarpowerworldonline.com/2014/02/eaton-releases-2mw-2-25mw-energystorage-solar-inverters/
8

https://www.mhi-global.com/products/detail/lithium_pro_sys_spec.html

10

http://www.ecoult.com/technology/ubersystem-configuration/

Page 7
! of !18

It is important to note that while Lead-acid batteries have 1/5 1/6 the energy density of Liion, the specific energy is around 1/3, and weight is the critical factor when it comes to
determining the number of containers needed. Batteries of most chemistries are relatively
heavy: there is enough space left in a fully loaded container to have a multiple doors and a
corridor for repairmen.
All in all, while the number of containers is a key determinant of infrastructure costs, it
can not be stated that the chemistry with the higher specific energy would be always
superior - individual solutions must be evaluated, because the different manufacturers pack
very differently.
Cost breakdown of battery energy storage system installations
Based on information from the numerous case studies available on similarly sized
projects 11, the battery cells themselves are not the only major cost, they actually account for
no more than 1/3 of the total project cost. There are two ways to look at BESS costs: one
time installation costs and lifetime costs. The final viability assessment is of course based on
the second one.

Estimated project cost breakdown

Assembly,
shipping,
installation, testing, &
commissioning
15%

Enclosure, HVAC,
control system,
fire suppression, etc.
20%

Batteries, racks,
bus bars,
connections/cables
40%

Power Conversion
System
25%

The figure is from a 3 MW / 750 kWh lead acid installation undertaken in Alaska in 2012
to integrate additional wind power into an island system by providing frequency response.
The breakdown details one time installation costs only. It is important to note that the
http://www.irena.org/DocumentDownloads/Publications/
IRENA_Battery_Storage_case_studies_2015.pdf
11

Page 8
! of !18

advertised 750 kWh is the so called usable capacity: this system was designed with 20%
DOD cycling in mind, therefore the actual installed capacity is ~3750 kWh. The total
investment cost of this BESS was $3 million, therefore according to the above cost
breakdown one 500 MWh battery block cost around $150 000. The used technology in this
application was (now defunct) Xtreme Powers Powercell, which is an advanced lead acid
chemistry, comparable to Ecoults UltraBattery.
The running costs are relatively small for BESSs, but not negligible. For lead acid
batteries, the temperature must be kept in the ideal range to maximize lifetime, and apt
ventilation must be provided to avoid hydrogen buildup. The cells must be monitored and
periodically refilled with distilled water. Advanced lead acid chemistries avoid some of these
issues but still require refresh cycles. In case of a Li-ion based system, running costs may
be lower.
If the battery is regularly cycled, the roundtrip efficiency also drives costs. For example
when providing frequency response, the net load may be small, but the losses that occur on
fluctuation discharge the battery, and the cost of charging is on the battery operator. The
distribution of the regulatory signal can be very well approximated with a bell curve, and thus
the energy traffic can be estimated per unit power committed to regulation. The losses would
be proportional to the traffic.

Thinking inside the box

Market research: the building blocks of BESSs


it may look very
nice onbattery
a pie diagram,
many costs are impossible to divide in
A While
containerised
Li-ion
system
practice. When one is designing a MW-scale BESS for a particular application, he is not
shopping
foris Safts
battery
cells, storingrenewable
shelves,
interconnecting
fire of
suppression
typeand
consisting
17 parallel strings
Intensium Max
ready-to-install
generation
plants or direct cables,
systems,
but
rather
integrated
solutions
that
pack
these
into modules.
In our
objective
of 28 Synerion
modules
(24V).
containerised
energy
storage
system
connection
to the
grid.all
TheofIM20
is based
also incorporates
an electronic
designed
forthe
todays
electricity
grids,
10 parallel
comprising
of
finding
ideal
BESS
for blackonstart
plus strings,
some each
ancillary
services,Each
thestring
building
blocks are
Battery Management Module (BMM) that
and for the smart grids of tomorrow.
29 battery modules (24 V), delivering a
battery
modules of the 100 kWh
range, power converters in the
100 kW range, the
controls the temperature, voltage and
It provides a megawatt-level energy
nominal 700 V and a rated energy of 42,
interconnecting
cables
of
these
modules,
and
infrastructure
for
placing
these
modules.
In the
current of each
module and
estimates
storage solution which is readily scalable
58 and 62 kWh depending on the cell type.
to suit a wide variety of applications,
including installation in combination with

The IM20+ is a second generation


container developed for E and M cell

their state of charge and health.

From Li-ion cells to


containerised batteries

VL cell
Synerion module
Rack of Synerion module
Intensium Max - containerized solution

Page 9
! of !18
Intensium Max

multiple MW range, the battery cells and the power converters are usually not integrated,
and are available in separate container size solutions.
Unless one already has a suitable storage building on site, it is usually cheapest and
easiest to use container-integrated systems. Usually one can fit a 2 MW inverter into one 20
foot container. For advanced lead acid batteries, the same container can fit about 500 kWh
(Ecoult), and in case of li-ion, 500-1000 kWh, depending on the cell type. Most of the time 20
foot containers are used because of the weight limitations.
Inside, a container has many modules on racks. The example above is from Safts
Intensium Max 20 product range, which is a Li-ion using MW range energy storage solution,
housed in a 20 foot ISO container. The cells are organized into 24V modules that have a
total capacity of 1.5/2.1 kWh, depending on the cell type. These modules are the core
building blocks of the system. They are organized into strings, which are essentially modules
connected in series. Then the strings are connected in parallel to make up the whole
system. In an actual IM20 container, 29 modules are connected series in a string and 10
strings are in parallel. There are monitoring and control systems per module, per string, and
per container. As seen on the above picture, the container is far from filled, and provides
easy access for maintenance and replacement of the modules. AC/DC converters are not
included. The weight of such a system is around 15 tonnes, depending on the types of cells
used. Total capacity is 420 to 1000 kWh per container.
In case of lead acid technology, Ecoults same size containers are capable of storing 500
kWh, and their inner structure is very similar.

Apart from the local control systems, the containers are equipped with heating,
ventilation, air conditioning (HVAC), fire suppression, and connectors to join them into a
greater network. A container-based BESS does not need a lot of infrastructure: only a stable
flat surface to put the containers, and a connection to the grid - this is why in most cases
BESS installations are placed into existing substations. The container based design is key in
keeping shipping and assembly costs down. The control of the entire installation is usually
integrated into the inverter.
Therefore the actual cost breakdown would include the following well separable terms:
1. Battery containers

Page 10
! of 18
!

2. Control/power converter containers


3. Infrastructure for the container array: footing and interconnecting cables
4. Shipping, installation, commissioning
The third point will not be considered in this paper, because the power plant in mind
already has abundant suitable surfaces and the cost of interconnecting cables is negligible
compared to the whole cost. If I were to take the infrastructure cost into account, it would be
proportional to the number of containers, just like the shipping, installation, commissioning
costs.
As of the fourth point, these depend mostly on the location. Looking at the case study
from Kodiak Island, Alaska these costs were around $55000 per container. Since the
considered location in Hungary is connected to the road network and is actually just off of a
major highway, in the middle of Europe, this cost would be substantially lower: shipping of a
20 foot container on land costs no more than $2 per km12, and it may be assumed that the
distance will be less than 1500 km. A crane will be needed for unloading, that costs around
$700 a day, and will be able to place all the containers in one day. Commissioning the entire
system will be a few days work for a small group on engineers, I assume $200 per
container. Therefore I estimate the total cost of shipping, installation, commissioning at ~
$3500 per container. This relatively low cost favors the less compact solutions.
The price of inverters in the MW range is around $0.25/W, therefore $625,000 may be
estimated for 2.5 MW. This is independent of the selected chemistry. The inverter + control
system is expected to fit in one 20 foot container.
Finding the optimum
The goal of this paper is to decide whether replacing the diesel genset is economical and
to select the ideal capacity and chemistry if it is. The first simplification is taking the genset
out of the problem: the optimal BESS will be found and it may be compared to the diesel
later.
The aim is to optimize for returns: yearly profits over initial investment. The initial
investment was detailed in the previous section: it becomes a function of installed capacity
and chemistry trough the battery cost and the number of containers needed. For yearly
profits, a yearly cost and revenue function must be developed.
The yearly cost function is the sum of amortization, energy bills (HVAC and efficiencyrelated), and maintenance costs. The revenue is a yearly flat fee for providing black start
capability plus the income from whatever surplus capacity that can be sold for frequency
regulation. Peak shaving was ignored as an application because with such high performance
power converter, it would require an enormous (100 MWh range) storage to operate
economically. As of 2016, peak shaving is an economical application in isolated grids, but
not on such an enormous, well interconnected grid as UCTE.
The profit function, with respect to Hungarian prices and regulations
The revenue from providing black start capability is a flat daily fee, RBS. The contract is
awarded for one year of nonstop operation. In the last year in Hungary, three black start

http://www.containerhomeplans.org/2015/07/how-much-does-it-cost-to-transport-ashipping-container/
12

Page 11
! of 18
!

generators were chosen at fees ranging between 1.7 and 2 million HUF13, that is a yearly
revenue between 2.25 and 2.65 million US dollars each. In order to reliably provide the black
start service and adhere to the standards, CBS = 1 MWh of capacity must always be ready to
dispatch to satisfy the requirement.
The revenue from frequency regulation is more complicated. The transmission system
operator (TSO) pays per committed MW, on an hourly (or 15 minute) basis, and this fee may
vary within the day - but it is decided in advance in the form of quarterly contracts. In this
model, a flat hourly rate is supposed.
In case regulation is provided with an energy reservoir that limits the regulation providing
capacity, ENTSO-Es Network Code on Load-Frequency Control and Reserves regulates
that the primary regulation must be provided as long as the reservoir is not exhausted. The
plant shall be able to fully activate its FCR continuously for a time period of not less than 30
minutes and for an equivalent longer time period in case of Frequency Deviations smaller
than the FCR Full Activation Frequency Deviation and shall specify the limitations of the
energy reservoir in the Prequalification. It is also required that in case of depletion the
energy reservoir must be recovered as soon as possible, and in no more than 2 hours.
Taking all this into account, the revenue:
RFR = 24 * 365 * rate * min(P,(C-CBS)/ReqENTSO-E), where C is total capacity.
The amortization cost will be proportional to the traffic and will depend on what the DOD
was as the power flown trough. Taking this second criteria into account is complex and is
considered out of scope for now. The expected hourly traffic per MW committed may be
calculated from the distribution on the regulation signal - this is discussed in detail later.
Expected lifetime = Cycle life * C / ( Traffic per MW * min(P,(C-CBS)/ReqENTSO-E) )
When calculating the amortization cost, the inverter and the rest of the installation must
be depreciated at a different rate - while the lifetime of some batteries is just a few years, the
rest of the installation may long survive them. In my calculations I used 10 years for the
lifetime of the non-battery part.
CAM = Battery cost / Expected lifetime + (Installation price - battery cost) / 10
The energy cost is a sum of multiple components: HVAC, self discharge, and efficiencyrelated waste. While of course HVAC is also a function of the battery roundtrip efficiency, it is
complex - for simplicity I take a flat rate of CHVAC = $5000/year/container. The self discharge
of the batteries depend on the applied chemistry. In case of lead acid, this is about 5% per
month, so ~70 W of continuous load per 1 MWh of installed capacity, which costs about
$0.01 per hour or ~$90 a year.
For maintenance, I used a flat rate of CMT = $1000 per container.
CEFF = (1 - roundtrip)/2 * traffic * cost of energy
Profit per year = RBS + RFR - CAM - CHVAC - CEFF - CMT
This profit is then subject to 19% corporate tax in Hungary.
The Hungarian Ancillary Services Market
The Hungarian transmission system operator (TSO) MAVIR is procuring ancillary
services by regularly publishing requests for proposals. The services are standardized to
facilitate the processing of the proposals and to enhance transparency. Any certified
company may submit a proposal and the lowest offer in each category takes the deal.

http://www.mavir.hu/documents/10258/208856549/PUB_BS_UQ_2015_11_18.pdf/
ef84dd0f-ce52-45b1-98ab-382c202c2383
13

Page 12
! of 18
!

Black start services are contracted on a yearly basis, as well as voltage and reactive
power regulation. In case of primary, secondary and tertiary reserves, service providers have
to submit proposals each quarter. The companies may offer primary reserve in symmetric 1
MW units, meaning that they must be able to regulate up or down as well. Secondary and
tertiary regulation is each separated into up and down regulation, and may be contracted in
5 MW units. 200 MW of secondary up, 100 MW of secondary down, and 500 MW of tertiary
up regulation is contracted - tertiary down regulation is not procured at all. In case of
secondary and tertiary regulation, there is also an energy fee, but not so with primary
regulation.
While the details of the contracts are not public, information on the distribution and final
price of the individual procured services is available on MAVIRs website.
Based on the ENTSO-E Continental Europe Operation Handbook, Load-Frequency
Control and Performance chapter, Frequency Restoration Reserves (FRR or secondary
regulation) should have a few minute response time, maximized in 15, while Replacement
Reserves (RR or tertiary regulation) are required to be online in 30 minutes of activation
from the TSO. Therefore in these applications the main competitive advantage of battery
based energy storage - very fast ramping - is irrelevant, and they will not be considered in
this paper.
On the other hand, Frequency Containment Reserves (FCR or primary regulation) must
be online in a matter of seconds: the deployment time for 50 % or less of the total
PRIMARY CONTROL RESERVE is at most 15 seconds and from 50 % to 100 % the
maximum deployment time rises linearly to 30 seconds. Pricing is matched to the tougher
requirements: while in Q2 2016 the FRR availability fees in Hungary rarely exceeded $30,
the average hourly FCR availability fee was $140.14 Even this application is far from utilizing
the ramping capabilities of a BESS, because it must be slowed down to mach that of
conventional spinning reserves: primary control within the entire synchronous area follows
the same deployment times to facilitate control.
FCR is also more fitting to battery based solutions, because it is contracted in relatively
smaller, 1 MW units, and because the minimum duration for the capability of delivery for
primary control is 15 minutes. There is one unit operational in Schwerin, Germany, providing
black start capability and primary regulation with a 5 MW Li-ion backed BESS.
http://www.mavir.hu/documents/
10258/211010693/2016_Q2_PRIMER_Eredmnyhirdets.pdf/5b0c582b-93ec-4770a76b-68f534290ab3
14

Page 13
! of 18
!

The costs of providing Frequency Containment Reserve


The cost of the service will be a function of the amount of traffic related to the service:
this will be proportional to the amortization of the batteries and the energy lost on roundtrip
efficiency. Committing 1 MW to primary regulation does not mean that it would be constantly
used, rather that it is the absolute maximum that the TSO can expect from the plant. Each
regulator has a so called K-factor that describes just how much power ought to be
dispatched proportionally to the frequency deviation. For the entire ENTSO-E system, the
FCE capacity is 3000 MW in both directions, and it is fully dispatched if the frequency
deviation exceeds 200 mHz. This means a K-factor of 15 000 MW/Hz, and therefore a factor
of 5 MW/Hz for each individual MW committed. This is a bit further complicated by the 10
mHz dead band around 50.0 Hz in which no control action is undertaken.
To estimate the traffic load from FCR, I analyzed some data.15 The measurements are
from the second week of September, 2012, 1 mHz resolution, sampled every second - a
total of 604800 data points. As expected, the frequency has a 50 Hz mean normal
distribution, and the standard deviation as calculated from the data was 67.48 mHz. The line
frequency was in the dead band 38.6% of the time, so no FCR was dispatched at those
moments.

Taking this dead band into account, the average active power flow magnitude on the bus
of the BESS would be just 72 kW per MW of FCR committed. The battery wear depends on
the actual SOC of the BESS and the C rate - therefore I used a simpler approximation to
estimate this cost component. The battery wear cost is the amortization, which is installation
cost over expected lifetime.

15

http://www.mainsfrequency.com

Page 14
! of 18
!

It could be argued that the value of a lead acid battery that reached the end of life criteria
is far from worthless - here I assume for simplicity that the recovered value would be spent
on transportation and installation of the replacement battery.
Cycle life may be defined in many ways. The above equation suggests that I defined it as
the number of times one may deplete the battery from 100% to 0% and charge it back up
again before it reaching end of life, commonly defined at 80% of original capacity. This kind
of operation is however very very different from the actual load it will face during FCR,
therefore I used data from a research paper where the battery was cycled between 45% and
55% SOC, and simply divided the so measured cycle life by 10 - thus ending up with 1500
for advanced lead acid.16
As is is clearly visible on the distribution of the frequency above, the bell curve is slightly
skewed to the right as the average frequency of that period was actually 49.998 Hz.
Accordingly, the FCR would have to up regulate more of the time than down, therefore net
depleting the battery. This effect is not negligible, the net consumption over the entire week
would have been 1.34 MWh, a value comparable to the roundtrip inefficiency. It is actually
not so interesting what the weekly consumption is - it is expected to average out to zero on
the long term, but it incurs costs none the less, as the battery should always be kept around
50% SOC to provide regulation, therefore a 1.34 MWh change in momentary charge is not
acceptable. Some algorithm must be developed that would start charging/discharging the
battery if its average SOC for the past few hours moved away from 50%.
The roundtrip efficiency depends greatly on just what that particular roundtrip looks like in general, for lead acid, the lower the SOC, the better the charge acceptance, as long as it
is above the deep discharge range. (10-15% SOC.) Therefore its efficiency for cycling
between 30 and 50% is much higher than if the same battery is cycled between 85 and
100%. In the 50-65% range, 90% is achievable, including the losses on the inverter as well.
The resulting cost is that of that wasted energy:

Conforming to the standards of FCR


As it was mentioned earlier, the standards set forth by ENTSO-E require that an FCR
providing plant with a limited energy reservoir be able to provide continuous regulation in
either direction at full activation frequency for 15 minutes. After one such disturbance, the
reservoir must be restored as soon as possible, but in no more than 2 hours, and must be
able to do the same again if the frequency deviates in such ways.
Therefore the BESS must have committed power times 0.25 hours of dispatchable
energy in it, and capacity to store that much more, at any time during normal operation. In
order to calculate just how much capacity that requires, normal operation must be defined obviously, if the system had to up regulate nonstop at full power for 10 minutes, it is not
reasonable to expect that it will be able to up regulate 15 minutes more just because there
was a few minutes of time in between when no regulation was necessary.
Because the reservoir restoration time is also two hours, a reasonable definition is to say
that if the average line frequency was ~50 Hz in the last two hours, the operation is deemed
normal and the plant is supposed to be able to regulate at least 15 minutes in each direction
at full capacity. To quantify it, normal operation is when the average of the line frequency
over the last two hours was 50 Hz +- 2*Std.Dev. The standard deviation may be calculated
16

http://www.sandia.gov/batterytesting/docs/LifeCycleTestingEES.pdf

Page 15
! of 18
!

from the deviation of the one second resolution frequency measurements by simply diving it
by the square root of the number of seconds in two hours.

With this definition, 95.5% of the time operation will be deemed normal if the distribution
of the momentary frequency is indeed normal. However, because of the dead band effect,
the BESS will be in the optimal SOC range with a lot higher probability!
The range in which the SOC moves during normal operation must be added to 2 * 0.25 *
committed power to get the capacity with which the standards may be met. This range may
be estimated using the above frequency boundaries: if the average frequency was 50.00159
Hz over the last two hours and we neglect the dead band, 0.00159 * 5000 * 2 = 15.9 kWh of
energy must be stored in the BESS per MW committed FCR. Therefore 532 kWh must back
up 1 MW of regulation power.
Cost and revenue breakdowns
Having introduced the different cost and revenue components, below is a calculated cost
breakdown a 3 MW, 3 MWh system, the optimal size for both chemistries:
Installation costs
Item

LA

Li-Ion

500 kWh advanced LA battery container x6

900000$US

1500000$US

3 MW inverter (fits in two containers)

750000$US

750000$US

28000$US

28000$US

1678000$US

2278000$US

Shipping and commissioning 7 containers


Total

Yearly running costs


Item

LA

Li-Ion

193721$US

95977$US

75700$US

75700$US

274$US

274$US

Dissipated energy due to inefficiency

14198$US

14198$US

HVAC

40000$US

40000$US

8000$US

8000$US

331892$US

234149$US

Amortization of batteries
Inverter amortization
Self discharge

Maintenance
Total

The expected yearly revenue from operating black start service and frequency
containment reserve:

Page 16
! of 18
!

Item

Revenue

Black start yearly standby fee

2282909$US

Frequency regulation revenue

3798655$US

Total

6081564$US

For advanced lead acid, the above detailed costs and revenues add up to a yearly profit
of $5 749 671 before taxes, $4 657 234 after tax, a 278% margin. The same for Li-Ion is
slightly higher profit but just a 208% margin as the initial investment in substantially higher.
Weak points and future research
Currently the weakest part of this paper is the reliability of prices assumed for battery
prices, which is key in assessing profitability. Manufacturers are not releasing the prices of
their container sized products, and while there are a wide variety of levelized prices available
on the internet, they are either given for small (cell phone battery size, AA size, car battery
size) applications or for entire grid scale projects, including inverters, maintenance and/or
other undisclosed costs.
A Tesla Powerwall retails at $3 000, is indeed nothing else but a big li-ion battery with
some controls, and has a capacity of 6.4 kWh. It is safe to assume then that the same
technology scaled up to 500 kWh, with centralized
control and one big container for housing in
Running costs breakdown,
place of the plastic/metal frame of the Powerwall
would not increase the levelized cost of $469,
advanced LA, 3MW, 3MWh
making the price of a 500 kWh container $234
500. While I believe that the economies of scale
outweigh the additional costs of the 20 foot
2%
container (about $2 000) and HVAC, rounding
12%
up to $250 000 is a very safe estimate. Teslas
utility scale Powerpack solution which comes in
4%
100 kWh units ready for outside installation
comes at an undisclosed price but analysts
predict a $250/kWh levelized price target which
58%
is exactly half of what I assumed.17
23%
A 200 Ah (2.4 kWh) truck starter battery
retails around $20018, which would put the cost
of a 500 kWh unit at ~$42 000. In this case
however the battery control system, inside
connections, and housing must be added,
rounding it up to 50, maybe 60 thousand US
Amortization of batteries
dollars. In the calculations I used a price of $150
Inverter amortization
000, which I acquired from an installation cost
Self discharge
breakdown of a BESS plant. One reason for the
Dissipated energy due to inefficiency
difference is that that was advanced lead acid
HVAC
versus the quoted truck battery which is
Maintenance
conventional.
https://transportevolved.com/2015/05/06/tesla-motors-posts-q1-2015-losses-due-tostrong-dollar-high-capital-expenditures-hits-1000-carweek-model-s-production/
17

18

http://www.akkutop.hu/index.php?route=filter&filter=category%7C54/amperora%7C12

Page 17
! of 18
!

There are also differences between two batteries of the same general chemistry as
different factories will produce slightly different qualities - therefore the cycle life data is also
to be taken with a grain of salt. A much more accurate calculation would be possible with this
model given accurate data - the gathering of which is one of the ways this paper is to be
developed.
It would also be interesting to consider adding photovoltaic generation to the plant to
further exploit the capabilities of the inverter. Some PV capacity could be locally used to
cover the losses of the BESS and a bigger installation could further increase the reliability
and robustness of the black start capability.
Yet another way to go is mixing different technologies. Since in this application the BESS
is kept at 60-70% SOC at all times except for emergencies, it may be wise to utilize some
cheap-but-low-cycle-life technology for that 90% of the BESS that is rarely touched and
actively cycle some more expensive but also more durable technology in the remaining 10%.
Developing a model for optimizing such hybrid solutions would be a major work that could
yield interesting application-specific results. In order to make it more accurate, a more
detailed SOC-dependent battery wear model must also be implemented.
Gergely Marton
2016. 05. 16.

Page 18
! of 18
!

Vous aimerez peut-être aussi