tensorkelechi/arxiv-cstext · Datasets at Hugging Face

text stringlengths 0 89.3k
BSmin
iBSiBSmax
i 22
Constraint 18 is the operation equation in which param
eters ηchandηdisare charging and discharging efficiency
respectively Constraints 1922 are limits of charging and
discharging rates states of charge and sizes respectively
In this problem the battery size BS is chosen as our
design variable and the remaining operationrelated ones
as our control variables including charged and discharged
power of batteries PchPdis active and reactive power
output of generators PGQG AC and MTDC power flow
related variables csvin constraints eq 46 eq 14 and
the power through MCC PConvPLossPDC
To convert this problem into a graphbased model we
define each ACDC bus ACDC branch converter and
BESS as graph nodes The linking constraints of the edges
and objective terms in each node are derived based the
abovedescribed formulation We consider eight hours as our
problem time horizon The resulting graph is shown in Figure
3 with each color representing an hour The structure of the
graph shows the nodes with associated decision variables
connected in time sequential hours While selecting one of
the hours on the graph circled in black shows the linking
constraints among different components including the same
components of the system
IV R ESULTS
DC grid branch data is presented in Tables I The converter
droop control and efficiency parameters are k 002d 1
β 003 The batteries minimum BSmin and maximum
BSmax capacities are 20MWh and 120MWh respectively
and their initial and minimum states of charge at the last hour
are set to half of BSmin Batteries charge and discharge
efficiency are 08and11 respectively Nominal wind farm
WF outputs are set as 40MW and 50MW respectively
Nominal load and generator cost functions are taken from
the 9bus system data In each hour WF output nominal
load and fuel cost are multiplied by the factors provided in
Table II The problem is formulated in Julia programming
with packages JuMP andPlasmo solved by Ipopt and
Juniper on a laptop with i712800H CPU and 16GB ram
Fig 3 The graphbased codesign optimization problem visualization for
IEEE9 busMTDC use case Each color indicates a time snapshot of the
complete system described as a graph node for 8 Hours
TABLE I
DC G RIDBRANCH DATA
From To Rpu From To Rpu
1 2 00016 2 3 00048
1 4 00048 3 4 00042
A Single Objective Function
We first only consider objective function eq 2 for
minimizing total cost The results by setting each of the two
batteries with the same fixed sizes ranging from 20MWh to
120MWh in 10 MWh increments are shown in Figure 4
Fig 4 Minimum total system costs with different battery sizes
By setting battery sizes as decision variables total cost is
minimized with an optimal solution of 265 MWh and 932
MWh The time for solving this problem is about 45 seconds
while solving all 11 problems with fixed battery sizes takes
almost 15 minutes in total
To better understand how the batteries help reduce total
cost we investigate in detail their hourly operations In
Figure 5 the total charging positivedischarging negative
power and state of charge of the batteries are shown by solid
lines by the subfigures respectively including results with
TABLE II
LOAD W INDFARM OUTPUT AND COST LEVELS
Hour 1 2 3 4 5 6 7 8
Load 09 11 125 14 155 13 115 1
WF Output 1 095 105 09 085 1 11 095
Fuel Cost 11 09 13 15 18 16 14 1411 fixed battery sizes and that of CCD by setting battery sizes
as decision variables The dashed lines are relative load and
generation cost levels in Table II The BS in legend stands
for Battery Size When load and generation costs are low
like in the first and last few hours the batteries are charged
These charged power is then used to satisfy demand when
load and cost are high By transferring load from peak to off
peak hours the total cost is reduced By choosing the best
battery sizes based on singleobjective control codesign our
model can balance the cost of battery installation operation
and power generation
Fig 5 Different battery sizes and relative costload levels in each hour
a Battery chargingdischarging operation b Battery state of charge
Since batteries transfer load from peak to offpeak time to
reduce cost it is not difficult to imagine that different load
levels may require different battery sizes We calculate the
total cost of the system based on these 11 fixed battery sizes
considering demand levels between 98 and 104 of the
original nominal demand The results are shown in Figure 6
As the load increases the best battery sizes to have minimum
total cost also increase On one hand higher demand may
need larger battery capacities for transferring load from peak
to offpeak hours On the other hand higher generation cost
resulting from larger load allows more budgets for battery
installation and operation
Fig 6 Total system cost of different nominal demand and battery sizes
B Multiple Objective Functions
In this subsection we minimize both objective functions
eq 2 and eq 3 together We first simply adopt the
linear scalarization approach to find the Pareto front ie

BSmin

iBSiBSmax

i 22

Constraint 18 is the operation equation in which param

eters ηchandηdisare charging and discharging efficiency

respectively Constraints 1922 are limits of charging and

discharging rates states of charge and sizes respectively

In this problem the battery size BS is chosen as our

design variable and the remaining operationrelated ones

as our control variables including charged and discharged

power of batteries PchPdis active and reactive power

output of generators PGQG AC and MTDC power flow

related variables csvin constraints eq 46 eq 14 and

the power through MCC PConvPLossPDC

To convert this problem into a graphbased model we

define each ACDC bus ACDC branch converter and

BESS as graph nodes The linking constraints of the edges

and objective terms in each node are derived based the

abovedescribed formulation We consider eight hours as our

problem time horizon The resulting graph is shown in Figure

3 with each color representing an hour The structure of the

graph shows the nodes with associated decision variables

connected in time sequential hours While selecting one of

the hours on the graph circled in black shows the linking

constraints among different components including the same

components of the system

IV R ESULTS

DC grid branch data is presented in Tables I The converter

droop control and efficiency parameters are k 002d 1

β 003 The batteries minimum BSmin and maximum

BSmax capacities are 20MWh and 120MWh respectively

and their initial and minimum states of charge at the last hour

are set to half of BSmin Batteries charge and discharge

efficiency are 08and11 respectively Nominal wind farm

WF outputs are set as 40MW and 50MW respectively

Nominal load and generator cost functions are taken from

the 9bus system data In each hour WF output nominal

load and fuel cost are multiplied by the factors provided in

Table II The problem is formulated in Julia programming

with packages JuMP andPlasmo solved by Ipopt and

Juniper on a laptop with i712800H CPU and 16GB ram

Fig 3 The graphbased codesign optimization problem visualization for

IEEE9 busMTDC use case Each color indicates a time snapshot of the

complete system described as a graph node for 8 Hours

TABLE I

DC G RIDBRANCH DATA

From To Rpu From To Rpu

1 2 00016 2 3 00048

1 4 00048 3 4 00042

A Single Objective Function

We first only consider objective function eq 2 for

minimizing total cost The results by setting each of the two

batteries with the same fixed sizes ranging from 20MWh to

120MWh in 10 MWh increments are shown in Figure 4

Fig 4 Minimum total system costs with different battery sizes

By setting battery sizes as decision variables total cost is

minimized with an optimal solution of 265 MWh and 932

MWh The time for solving this problem is about 45 seconds

while solving all 11 problems with fixed battery sizes takes

almost 15 minutes in total

To better understand how the batteries help reduce total

cost we investigate in detail their hourly operations In

Figure 5 the total charging positivedischarging negative

power and state of charge of the batteries are shown by solid

lines by the subfigures respectively including results with

TABLE II

LOAD W INDFARM OUTPUT AND COST LEVELS

Hour 1 2 3 4 5 6 7 8

Load 09 11 125 14 155 13 115 1

WF Output 1 095 105 09 085 1 11 095

Fuel Cost 11 09 13 15 18 16 14 1411 fixed battery sizes and that of CCD by setting battery sizes

as decision variables The dashed lines are relative load and

generation cost levels in Table II The BS in legend stands

for Battery Size When load and generation costs are low

like in the first and last few hours the batteries are charged

These charged power is then used to satisfy demand when

load and cost are high By transferring load from peak to off

peak hours the total cost is reduced By choosing the best

battery sizes based on singleobjective control codesign our

model can balance the cost of battery installation operation

and power generation

Fig 5 Different battery sizes and relative costload levels in each hour

a Battery chargingdischarging operation b Battery state of charge

Since batteries transfer load from peak to offpeak time to

reduce cost it is not difficult to imagine that different load

levels may require different battery sizes We calculate the

total cost of the system based on these 11 fixed battery sizes

considering demand levels between 98 and 104 of the

original nominal demand The results are shown in Figure 6

As the load increases the best battery sizes to have minimum

total cost also increase On one hand higher demand may

need larger battery capacities for transferring load from peak

to offpeak hours On the other hand higher generation cost

resulting from larger load allows more budgets for battery

installation and operation

Fig 6 Total system cost of different nominal demand and battery sizes

B Multiple Objective Functions

In this subsection we minimize both objective functions

eq 2 and eq 3 together We first simply adopt the

linear scalarization approach to find the Pareto front ie