Sie sind auf Seite 1von 6

A MODEL FOR PLANNING OF DISTRIBUTED GENERATION IN THE

LOCAL TRANSMISION SYSTEM


Geir Warland and Michael Belsnes

Power Generation and Marked
Sintef Energy Research
N-7465 Trondheim, Norway


Abstract

This paper presents a model that combine the analysis of available power capacity from hydro power
plants with an analysis of the available transmission capacity in a local power grid. This model is able
to optimize the power plant discharges taking into account transmission constraints imposed by the
local transmission grid. The ability to include transmission grid constraints enables the model to
analyze problems such as: rehabilitation and maintenance planning in the local transmission grid,
simulation of the effect of introducing new (renewable) energy and distributed generation.




1 INTRODUCTION
The Nordic countries, Norway, Sweden, Finland and
Denmark today share a common spot market for
electrical power. Deregulation and market competition
were introduced in the electricity market in Norway in
1991, in Sweden in 1996, while Finland and Denmark
came later. Transmission constraints will, however, at
times still cause differences in spot prices in these
countries.
In the present market, and especially in Norway,
generation is to a great extent based on hydropower
(almost 100% in Norway). An important characteristic
of a hydro power system is that market prices may
vary greatly even in the space of a few months, or
even weeks, depending on variations in inflow in
addition to variations in consumption. Consequently,
in the time span to be considered when analyzing new
investments in the power system, there is great
uncertainty in market price.
A producer analyzing an investment in new power
production in Norway would have to consider, in
addition to the development of domestic demand and
hydro capacity, the following factors when forecasting
future prices:
The uncertainty about whether one or more
Swedish nuclear plants will be retired within the
foreseeable future. In a referendum Sweden
earlier decided to close all nuclear plants before
the year 2010. The retirement of a nuclear plant
would probably tend to raise market prices, how
much depending on what it is replaced with.
The probability of an increased capacity for
exchange with continental Europe through new
undersea cables. Increased cable capacity would
probably increase demand for peak power, but
reduce annual variations in price due to variable
precipitation. This is because the continental
European power system is largely thermal based,
with ample capacity for selling energy in off-peak
hours.
Increasing thermal generation capacity in
Norway. Two natural gas fired plants are already
in the planning stage. The effect of introducing
these plants into the Norwegian market is
somewhat uncertain. They will probably be base-
load plants, possibly contractually tied to
exporting generated electricity.
Within the next few years, variations in price within
the day and week are expected to increase. This is in
large part due to the increasing peak load demand tied
to new undersea cables, but increasing domestic
demand combined with a lack of investment in new
capacity the last few years will also contribute to these
expectations. This will of course affect which projects
it will be profitable to invest in.
Models commonly in use up to now, however, have
not been capable of properly evaluating the value of
either peak capacity or distributed production; the
main concerns up to now have been tied to annual and
seasonal variations in inflow and market price, and
this is what present models mainly are designed to
evaluate the effects of. The need for evaluating
expected revenues from hydro plants in a market
where prices might vary just as much within the span
of a few hours as they do in the course of weeks or
months, means newer models have to be developed.
There is also an increasing need for evaluating
expected revenues from investments in distributed
power production and in investments in the local
transmission system.
The next chapter gives a brief overview of generation
expansion and refurbishment planning. Chapter 3
describes a model for performing static analyses, with
ability to handle market prices with price variations
within the span of hours, and performing analyzes of
distributed generation. Chapter 4 presents a case study
and conclusions are found in chapter 5.
2 PROBLEM OVERVIEW
Generation expansion or refurbishment planning
should ideally solve a three-dimensional investment
problem for a given planning period:
Type and location of the new generation
equipment.
Capacity expansion.
Time of investment.
The planning period should at least be equal to the
lifetime of the equipment. The generation expansion
problem is affected by several external variables,
which are uncertain:
Water inflow to the reservoirs.
Future market prices.
Investment costs.
The properties of the new equipment.
The values of the uncertain variables are coupled in
time. For example, the probability of the market price
for a given week in year 2002 is strongly dependent
on the market price in the previous week. The
operation of the production system can also depend on
the planned investments. The generation expansion
problem should therefore ideally be formulated as an
integrated stochastic dynamic optimization problem,
which includes both the investment costs and the
operational scheduling.
For many practical purposes it is not possible to solve
this optimization problem by current optimization
methods. The most commonly used method by
producers in Norway decomposes the generation
expansion problem into two subproblems:
Calculation of operational revenues for a given
production system by simulating system
operation. The time resolution used in these
calculations is typically one week.
Calculation of present value for a given system
including the investment costs.
The uncertain variables, which are independent of the
values the previous year, e.g. inflow, are modeled as
stochastic variables in the system simulation model.
The simulated expected operational revenues are used
for calculating the present value. Normally the
uncertainty in these variables is modeled by using
statistics, based on historical observations, where each
year is assumed to have the same probability.
The uncertain variables whose values are strongly
dependent on the value the previous year are fixed at a
constant level for a given period. The uncertainties in
these variables are included manually by calculating
operational revenues for different values, i.e. by
assuming various scenarios.
The market price is usually modeled using a
combination of these methods. On the one hand,
market price is strongly dependent on precipitation
and is therefore modeled by the same methods as
inflow. Forecasts for market price are obtained by
using a model such as the EFI Multi-area Production
Scheduling (EMPS) model [1] for simulating optimal
operation of the global market (Norway, Sweden,
Finland and Denmark with their ties to continental
Europe and Russia). On the other hand, market price
is coupled to external variables such as oil price or
national and international policies, e.g. regarding
environmental taxes. These uncertainties are typically
modeled by calculating the operational revenues for
different levels of these values, for example a high,
medium and low level.
This type of investment analysis is often referred to as
static analysis, where the results are valid for a given
future.
3 A NEW MODEL FOR PLANNING OF
DISTRIBUTED GENERATION
This chapter presents a model for evaluating expected
revenues from investments in a hydropower
dominated system in a market regime by simulating
(as closely as possible) optimal system operation is
presented. The new model is called Optimal Regional
Planning (ORP), and is an extension of an existing
long term scheduling model, sometimes referred to as
an Extended Single-Reservoir Model [2]. System
operation is calculated in two stages:
1. An optimal strategy is computed, using stochastic
dynamic programming to calculate expected
incremental water costs for an aggregate energy
model of the hydro system [3],[4]. The time
resolution is one week.
2. Hydro system operation is simulated for a number
of price and inflow scenarios, using a detailed
hydro model to evaluate revenues. The time
resolution is a number of hours, 24 at most.
At present our model implies the following
assumptions:
The system owner is a price taker, i.e. the market
price is not influenced by short -term variations in
the owners generation. Any effects an expansion
might have on long-term price development
could, however, be accounted for by recalculating
the price forecast with the expanded system
included in the price forecasting model.
System operation reflects a risk neutral strategy.
A strategy averse to risk can to a certain extent be
accounted for by introduction of a penalty
function [5].
For each inflow scenario there is a corresponding
market price scenario. Because market price for one
week is very much dependent on the previous weeks
price, an autoregressive model is used to model future
market prices for computing hydro system strategy
[5].
3.1 Simulating system operation
Hydro generation is simulated for a number of
price/inflow scenarios. This involves a weekly
decision process in three stages:
1. Aggregate hydro generation is found in a market
clearance process where hydro costs are
represented by the incremental water costs,
calculated in advance.
2. Aggregate generation is allotted to available plants
in a rule-based reservoir drawdown model.
3. An optimal schedule for the week is recalculated
with a finer time resolution.
In the first two stages the time resolution is one week,
but with a limited number of accumulated price
intervals within the week (usually four). In stage three
the time increment is up to 24 hours; usually the day is
divided in two or three time intervals the first five
weekdays, while the weekend consists of two 24-hour
intervals. The decision process is illustrated in fig. 1.
Fig. 1. Weekly simulation of hydro system operation
An explanation to fig. 1. :
A Optimize aggregate generation
The optimum aggregate hydro generation is
calculated, given a market price (which may vary
within the week). Hydro power is modeled with an
aggregate reservoir, and an aggregate plant with a
piecewise linear efficiency rate as well as a
piecewise constant operational cost. This cost is
added to the incremental cost of stored energy.
B Reservoir drawdown
A detailed reservoir drawdown model affords the
distribution of aggregate generation among
available plants, and thus the distribution of stored
energy among available reservoirs, according to a
rule-based strategy. When detecting an active
constraint which increases costs of generation by
more than 5%, the model exits temporarily B for an
Optimize aggregate generation
A
Reservoir drawdown
B
Update aggregate model
C
Model
modified?
Short term optimization
D
Last week,
last year?
Finished
Start
Next week
Yes
No
update of the aggregate hydro plant description in
C, and recalculation of optimum generation in A,
before returning to B for continuing the reservoir
drawdown process.
C Update aggregate model
Upon exiting the reservoir drawdown model, the
aggregate hydro model is always updated with the
latest inflow, generation constraints, pumping and
plant efficiency. If this modifies the aggregate
model in any way, optimum generation is
recalculated, by returning to A.
D Short-term optimization
Using end reservoir storage from B as target
storage levels for each reservoir, an optimal
schedule is recalculated with an increased time
resolution using a linear programming algorithm.
The short -term optimization model, including the
modeling of the transmission network will be
further discussed in section 3.3.
3.2 The reservoir drawdown model
As stated earlier, aggregate hydro generation is
allotted to available plants in a rule-based reservoir
drawdown model. The hydropower system is
modeled in detail, based on standard plant/reservoir
modules as shown in fig. 2.
Fig. 2. A standard plant/reservoir module.
Properties, which may be attached to each plant/
reservoir module, include:
A reservoir, defined by its volume and
relationship between stored volume and elevation
above sea level.
A plant, defined by its discharge capacity and a
piecewise linear relationship between discharge
and generation.
Variable head.
Inflow (weekly, daily for short term optimization)
either to the reservoir or directly to the plant
(implying it cannot be stored).
Independent destinations for plant discharge,
bypass discharge and reservoir overflow.
Time variable constraints on reservoir contents
and water flow (plant or bypass discharge).
Plant discharge, which is limited by variations in
head.
The basic goal of the reservoir drawdown strategy is
to generate a specified amount of energy in the
cheapest possible way, but this is done without any
formal optimization. The basic goal is sought fulfilled
by:
1. Seeking to minimize the risk of overflow during
that part of the year when inflow is greater than
discharge.
2. Seeking to avoid loss of head and power capacity
caused by emptying reservoirs too early during that
part of the year when discharge is greater than
inflow.
Despite the lack of formal optimization, it has fared
surprisingly well in head to head comparisons to a
stochastic optimization model.
3.3 The short-term optimization model
In the aftermath of the reservoir drawdown model, an
optimal schedule for the week is recalculated, by
solving a deterministic problem using a linear
programming algorithm. The reservoir drawdown
model has already come up with a reasonably good
schedule, but with a time resolution that does not
satisfy our needs. So the only results that are kept
from the reservoir drawdown model are end-of-the-
week stored volume for each reservoir.
The optimization goal is to maximize revenues from
sales of electrical energy for a week, assuming:
A fixed market price, which may however vary
from time interval to time interval within the
week. All generated power is sold at market
price.
A target level for each reservoir equal to the end-
of-the-week storage from the reservoir drawdown
model. Deviation from target reservoir level is
allowed, but entails a penalty.
Failure to meet minimum constraints on water
flow, are penalized economically.
The hydro system is modeled in exactly the same
detail as in the reservoir drawdown model. The main
difference is the finer time resolution and the use of
Local inflow
Uncontrollable inflow
Reservoir
Plant
discharge
Bypass
discharge
Overflow
Power plant
daily inflow statistics. Other points, which should be
noted, are discussed in the following paragraphs.
The model uses the results for end-of-week stored
volume for each reservoir, as reservoir target values
for a local linear deterministic optimisation of plant
discharges within the week. Using a linear
deterministic model results in a large degree of
freedom with respect to the types of problems that can
be modelled and analysed within the week.
Given target values for end-of-week storage volumes
for the reservoirs, and prices of electric power within
the week, optimal plant discharges are found for the
given week. This is repeated for each week, for all the
historical inflow values (or years). The process is
repeated for each historical inflow year.
The model allows an optimal plant discharge that
takes advantage of within day price-variations. Since,
also day-values of water inflow are used in the
simulations, this model also gives good results, in
terms of more realistic plant discharges and revenue
potential, in poorly regulated hydro-systems.
The concept is further enhanced, by adding additional
linear constraints when base-case or post-contingency
overloads are detected. In order to determine
overloads, a security constrained optimal power flow
(SCOPF) is used, using the optimal power schedule
found so far as input. The objective in this SCOPF is
to find the minimum redispatch required to avoid the
detected base-case or post-contingency overloads. The
results are used to add new constraints to the linear
optimization problem. New constraints are only added
when overloads are detected. By monitoring base-case
overloads as well as post-contingency overloads and
adding new constraints for each new overload, allows
the calculation of production strategies that can satisfy
security criteria, such the traditional N-1 criterion.
Introducing power grid constraints allows an increased
complexity in the types of problems that can be
analysed, especially with respect to distributed power
production. An example of a hydropower system and
the associated power grid is given in fig. 3.
3.4 Possible areas of application
The presented model can be used to analyze several
types of problems. Some of the possible areas of
application for this model can be listed as follows:
Fig. 3. Hydropower system and power grid
Generation expansion or refurbishment planning.
Rehabilitation and maintenance planning.
Planning of expansions in the electric power grid.
A supplement to the discharge planning given by
Extended Single-Reservoir Model [2].
Simulation of new wind power plants.
Simulations and analyses for distributed power
production.
4 CASE-STUDY
In this chapter a case study is presented in order to
demonstrate the use of the ORP-model. The model is
used to analyze the effect of introducing a wind farm
in a local area. A price forecast has been obtained
from the EMPS model, with demand modeled in four
intervals each week (peak load, mean load, off peak
and weekend). The ORP-model is used to simulate
before and after the introduction of a 50 MW wind
farm at bus B110, see fig. 4.
4.1 System data
The local power system has 10 reservoirs and 7 hydro
power plants. The total installed capacity in the area is
approximately 1150 MW and a total load is 875 MW
(with a local load of 150 MW at bus B110).
To model the wind farm, a module as in fig. 2. is used
with the wind energy modeled as uncontrollable
inflow. Fig. 4. shows the configuration of the
hydraulic system and the local transmission network.
4.2 Results
Prior to the introduction of a wind park, there is a
local bottleneck. Limited transfer capacity on the
> 132 kV
H27 H25 H21 H23 H22
B101 B110 B108 B102
H28 B114 H26 B116 B123 H24
B103 B121

H<i>: plant/reservoir module number i H29 B113 B109
: Hydropower system
B<i>: Bus bar (solid vertical lines) number i
: Transmission lines (from bus B<i> to B<j>) H30 B118 B108

Fig. 4. Local hydropower system with local transmission grid.
transmission line B101-B110, due to security
constraints (N-1-criterion), makes it necessary to
redistribute power production. With an outage of line
B110-B113, the transfer on line B101-B110 is too
high, partially due to the high local load at bus 110.
The power production at the locations H29 and H30 is
increased, resulting in a power flow that satisfies the
security constraints (the N-1 criterion).
Table 1. Average yearly, simulated results.


Generation
GWh
byp.+
overfl
av. hr.
N-1
Economy
MNOK
1)

Before
After
2308
2457(145)
130.6
129.9
17,7
16,0
626
667
1) 9.30 NOK = 1US$
The introduction of wind energy gives the results
shown in table 1. Power generation increases by 149
GWh, of which the wind farm only contributes with
145 GWh. The bypass and overflow are reduced with
0.7 GWh. This shows that by introducing wind power,
a more favorable despatch can be found for the
hydropower system. The table also shows that, even
though security constraints still persist, the average
number of hours per week has been slightly reduced,
from 17.7 to 16 hours per week.
5 CONCLUSION
The paper presents a model able to evaluate
consequences of investments in hydropower systems,
as well as in distributed power production. In the
model both the hydraulic system, as well as local
transmission grid are included. A model of this type
could be of great value both to potential investors and to
authorities.
6 REFERENCES
[1] Botnen, O.J., Johannesen A., Haugstad A., Kroken
S., Frystein O. 1992. Modelling of Hydropower
Scheduling in a National/ International Context,
proceedings from Hydropower 92, Lillehammer,
Norway, June 1992. Rotterdam: Balkema.
[2] Flatab N., Olaussen E., Hornnes K., Haugstad A.,
Johannesen A., Nyland S., 1988. EFI Technical
Report No. 3483: EFIs Models for Hydro
Scheduling. Trondheim: EFI.
[3] Stage S., Larsson Y. 1961. Incremental cost of
water power, AIEE Transactions (Power Apparatus
and Systems), August 1961.
[4] Gjelsvik A., Rtting T. A., Rynstrand J. 1992.
Long-term Scheduling of Hydro-Thermal Power
Systems, proceedings from Hydropower 92,
Lillehammer, Norway, June 1992. Rotterdam:
Balkema.
[5] Wangensteen I., Mo B., Haugstad A. 1995. Hydro
Generation Planning in a Deregulated Electricity
Market, proceedings from Hydropower Into the
Next Century, Barcelona, Spain, June 1995. Sutton,
UK: Aqua-Media International.
[6] Haugstad, A., Mo, B., Belsnes, M., Evaluating
hydro expansion or refurbishment in a deregulated
electricity market, Hydropower97.

Das könnte Ihnen auch gefallen