Sioshansi F.P. Smart Grid: Integrating Renewable, Distributed & Efficient Energy
Подождите немного. Документ загружается.
than is being consumed. As a result, there will be increasing
demand for the ability to deploy increased load instead of
only shedding wind power or turning a power plant off.
A recent case in point is Bonneville Power Authority's (BPA)
experience in dealing with over-generation events in the first
half of 2011. BPA is the transmission owner and operator for
75% of the high-voltage transmission lines in the Pacific
Northwest. Approximately 89% of BPA's 18.3 GW peak
capacity resources are hydropower (31 federal dams), 6.3%
are nuclear (1 nuclear power plant), and the remaining 4.6%
are primarily wind power.
37
But because BPA is the regional
transmission owner, it is also responsible for the distribution
of an additional 27.3 GW of peak capacity, of which 5.9 GW
is coal and 8.5 GW is natural gas.
37
Source: BPA [12].
Over the last five years, BPA has seen a rapid growth in wind
power within its balancing area (Figure 10.9), and BPA
estimates the installed wind power capacity in its balancing
area will double in the next few years.
38
The large
penetration of wind power in BPA (>30% of installed
capacity in BPA's balancing area providing 13% of BPA's
energy in 2010) has resulted in its heavy reliance on hydro
generators to provide balancing services. Hydropower is an
excellent resource to pair with wind generation due to
hydropower's fast response times, ramp rates, and lack of
CO
2
, NO
x
, and SO
x
emissions. But BPA's hydropower
facilities have operating constraints to ensure the health of
multiple flora and fauna species within the river systems in
Oregon and Washington. These hydropower operating
constraints are similar in nature, but more rigid, than emission
521
limitations of fossil-fuel power plants. These operating
constraints are flow constraints and river chemistry
constraints; thus BPA's dams have minimum and maximum
flow rates as well as dissolved oxygen and nitrogen levels
they must maintain for the river system downstream of their
discharge. As BPA has experienced over the last few years,
these environmental constraints have increased the
operational challenges of integrating wind power by reducing
the capability of BPA's hydropower facilities to provide
balancing services and increasing the occurrence of
over-generation events.
38
Source: BPA [13].
Figure 10.9
522
The growth of wind power capacity in BPA's balancing area.
Source: BPA [13]
The increased frequency of over-generation events led BPA
to formalize an environmental redispatch policy that provides
BPA with a formal plan for curtailing generation in its
system. When over-generation conditions are imminent, BPA
will first redispatch thermal generators to minimum power
levels that generators are capable of or are necessary for
reliability reasons. If this redispatch does not alleviate the
threat of an over-generation event, BPA will then curtail wind
generation on the system in order to allow its hydropower
facilities to meet their environmental constraints.
39
39
Ibid.
Current demand response programs would not have helped
BPA alleviate over-generation events as curtailing load would
only have worsened the conditions. Instead, it would have
been beneficial if BPA could have dispatched customers to
increase the load in their balancing area. This capability does
not exist except with customers enrolled in real-time pricing
programs. RTP responses to reduced prices, however, are
relatively small for most customers. Thus system operators in
the coming years will look to develop more advanced flexible
load control programs that can provide both up and down
flexible reserve services.
In order to effectively deploy demand response as an ancillary
service resource, a system first needs to determine the nature
and size of the flexibility services it would need to integrate
significant penetrations of variable renewable generation.
From this it can then employ an optimal inventory algorithm
523
to determine the amount of load participation needed and the
optimal manner in which it should be deployed. The concepts
and design of an optimal inventory algorithm for load control
products are discussed in the next section.
Structuring Load Control Products for
Intermittent Resources
To deploy demand response products for ancillary services,
system operators need to translate the total capacity and
energy ancillary service needs into a number of contracts for
specified blocks of energy to be delivered by participating
load. The implementation of demand response for the
provision of regulation and load-following is to some extent
similar to programs designed to meet critical peak conditions
by enrolling participating loads that are willing to be
curtailed.
40
Critical peak programs of such a nature are
generally directly linked to the seasonal characteristics of
peak demand on a given system. For example, if the control
area demand peaks in the summer, curtailable load programs
are implemented only for summer months, allowing the
system operator to exercise a limited number of curtailment
calls over the summer period. In contrast to critical peak
programs, ancillary service needs persist across all months of
the year. As a result, load control products must allow for
monthly differences in the number of participants and amount
of energy contracted.
40
An extensive treatment of design and implementation of curtailable load
contracts can be found in Oren and Smith [14].
Intermittent generation increases the seasonality and
variability of ancillary service needs. The seasonality of wind
speed and, by extension, wind generation output can be
524
observed across various time-scales—from intra-daily and
weekly cycles, to multi-year phenomena such as El Nino. The
variability of wind, as it pertains to the need for regulation
and load-following, is largely affected by short-term weather
patterns. The results reported by wind integration studies
establish the quantity of regulation and load-following that
would be needed to maintain reliable system operations while
accommodating increasing penetrations of renewable
generation. A portion of those ancillary service needs can be
contracted for provision via demand response with monthly
differences that recognize the seasonal patterns in output and
variability of renewable generation.
The system operator will need to contract for a sufficient
amount of energy deliveries to ensure that the system will be
able to meet any actual outcome of MWh energy of ancillary
service needs with a given probability (e.g., 95% likelihood of
meeting the need). The total amount of energy deliveries has
to be decomposed into interruptible contracts with a
three-dimensional property. Each contract should be designed
to represent (1) a block of MW quantity; (2) the fixed number
of contiguous hours at a time over which the quantity can be
called upon; and (3) the maximum annual number of times a
multi-hour delivery block can be called. Consequently, the
total size of the contracted controllable load energy will be
represented by the collection of all contracts (defined by a
block of capacity that can be exercised up to a maximum
number of sequential hours up to a maximum number of
times per year).
For example, a single contract could specify that the customer
can be called to provide one kW for no more than five
contiguous hours at a time and can be called to do so at most
525
10 times per year. This contract, therefore, provides at
maximum 50 kWh per year. It is possible, however, to design
two contracts that allow for the same maximum amount of
total energy per year, but in reality differ in how they can be
exercised by the system operator. In the preceding example,
the customer agrees to interruptions over no more than 50
hours per year with no single interruption lasting more than 5
hours. Alternatively, a contract may be designed to require a
maximum of 50 hours of interruptions without a limitation on
the length of each interruption. The latter contract provides
more value to the system operator and requires fewer unique
program participants. This difference arises from two
underlying challenges.
First, the system operator faces an inventory
problem—enough total energy needs to be contracted from
individual program participants to meet a designated portion
of overall system ancillary service needs. For example, if the
average energy need is 10 MWh with a standard deviation of
0.5 MWh, then a total of 11 MWh would need to be
contracted for delivery.
41
If each contract was worth 50 kWh
per year, then one potential solution would be to contract with
11 MWh/50 kWh = 220 participants (each agreeing to the 50
kWh contract described above).
41
That is, an amount equal to the mean (10 MWh) plus at least two
standard deviations (2 × 0.5 MWh).
There is, however, a second challenge—the system operator
is also faced with a “stacking” problem. While the total
contracted quantity might satisfy the identified total energy
need for ancillary services, if the contract limits each program
call to a limited number of contiguous hours (e.g., no more
526
than 5 contiguous hours per each of 10 annual calls), then the
system operator might run out of energy blocks that can be
“stacked” to meet a volatile spike in ancillary service needs
because some calls were already exercised at less than the
5-hour limit. When a contract is structured as a “call” contract
rather than a “total energy” contract, each time a call is used
for less than the maximum allowed amount of hours, the
energy from the unused hours is “lost.” To address this
complication, the initial solution of 220 participants could be
relaxed upward using a general rule-of-thumb informed by
historical experience.
42
42
For example, the system operator might enroll an additional 37
participants (~365 days/10 MWh) if the contract is of the “call” type.
There are several dimensions to establishing a pricing
mechanism for the ancillary service demand response
program. Continuing the preceding illustrative example, two
principal contract designs can be considered. Program
participants can be offered a “total energy contract”—in
essence it commits each customer to providing a total of 50
kWh of energy annually, which can be called over 50 hours
without a limit on the duration of each program period.
43
On
the other hand, a “call contract” might set forth a different
condition—each customer can be called no more than 10
times per year, with each call lasting for no more than 5
contiguous hours. Estimating the value provided by
participating loads in each of the two contract scenarios has to
rely on measures of avoided cost and capacity value.
43
This allows for the possibility that the operator might ask a participant to
respond to load control interruptions for a number of contiguous hours.
527
A critical peak pricing (CPP) demand response program
developed by Southern California Edison (SCE) provides a
useful framework in the present context.
44
Under this CPP
program, the utility may call a maximum of 15 critical events
per year. Demand response participants receive notification
the day before they are likely to be called upon.
45
In addition
to the limited number of calls per year, each event can only
last a maximum number of hours. Under these conditions, the
capacity value of the load resources is estimated as the
product of their expected deliverable capacity and the avoided
cost of capacity. The avoided cost is based on the capacity
value of a combustion turbine, but must be derated to capture
the reduction in value due to the lack of availability of the
load resource all year round. To estimate the derating factor,
the utility can run simulations in which participating load is
dispatched to meet critical system needs. Similarly, for
demand response providing ancillary services, the simulations
will model the percentage of times the load resources were
available to meet regulation and load-following needs.
Consequently, the derate factor can be decreased, thus
increasing the capacity value, by either offering contracts
with greater number of calls, longer call durations, or
“energy” type stipulations. Given this framework, an
“energy” type contract for 50 kWh as defined in the
illustrative example earlier will be more valuable than a “call”
type contract for the same amount of energy.
44
Chapter 23 by Braithwait and Hansen discusses CPP rates and empirical
outcomes of those programs in California.
45
This discussion follows closely [15].
528
Conclusions
The primary system and social benefits of using load
resources instead of generators for ancillary services are
increased portfolio diversity, reduced fuel and O&M
generator costs, reduced customer costs, and lower emissions.
Introducing wind into a system increases a system's
variability and leads to reduced utilization of conventional
fossil-fuel generators and as a result reduced revenues. Unless
other capacity and energy sources are introduced into a
market, prices will rise as conventional generators seek to
recover their lost revenues. Demand response is a perfect
complement to the increased penetration of wind energy
because demand response programs can provide capacity and
energy in limited quantities.
Fuel costs can be avoided by calling upon load resources to
integrate variable renewable resources instead of conventional
fossil-fuel generators when the economics to do so are
favorable. O&M costs are also reduced because maintenance
schedules for fossil-fuel power plants are determined by how
much cycling a plant does through its capacity range. By
calling on conventional generators less for ancillary services,
generators are able to extend the amount of time between
periods when they are overhauled and repaired. Customer
costs can be reduced because they will receive revenue
payments for capacity and energy as participating load
resources. Finally, carbon dioxide, nitrogen oxides, and sulfur
dioxide emissions are also reduced when load resources are
substituted for fossil-fuel generators due to decreased fuel
usage as well as avoided efficiency declines that occur when a
fossil-fuel power plant produces power below its specified
nameplate capacity.
46
529
46
For more information, see Katzenstein and Apt [16] or Denny and
O’Malley [17].
References
[1] Database for State Incentives for Renewables & Efficiency
(DSIRE), http://www.dsireusa.org/documents/
summarymaps/RPS_map.pptx (accessed 17.03.11).
[2] EIA, Annual Energy Outlook 2011 Early Release
Overview. U.S. Energy Information Administration,
2008, 9.
[3] CAISO, Integration of Renewable Resources. California
Independent System Operator, November 2007−July
2011, http://www.caiso.com/Documents/
Integration-RenewableResourcesReport.pdf
[4] NYISO, NYISO Supplement and Errata to NYISO Annual
Report on Demand Response Programs. Federal Energy
Regulatory Commission. Filing; Docket No.
ER01-3001-000.
[5] NREL, Eastern Wind Integration and Transmission Study.
National Renewable Energy Laboratory, February 2001,
134–148, http://www.nrel.gov/wind/systemsintegration/
pdfs/2010/ewits_final_report.pdf.
[6] Chang, J.; Madjarov, K.; Baldick, R.; Alvarez, A.; Hanser,
P., Renewable Integration Model and Analysis, IEEE
Transmission and Distribution Conference and
Exposition (2010).
[7] R. Shankar, Low Load/Low Air Flow Optimum Control
Applications. EPRI. TR-111541, December 1998.
[8] Monitoring Analytics, PJM State of the Market Report
2010. Monitoring Analytics. March 3, 2011.
530