tensorkelechi/arxiv-cstext · Datasets at Hugging Face

text stringlengths 0 89.3k
8 considered the BESS sizing problem for an onshore wind
farm however the authors developed a sequential approach
where they first chose BESS size and then proposed a control
strategy To the best of the authors knowledge none of
the studies has considered simultaneously accounting for the
control operations of offshore wind farms while sizing BESS
Conventionally the design problem is solved first followed
by operation control optimizations However many studies
915 have shown that such a sequential approach re
sults in suboptimal system performance Control codesign
CCD is a control system design approach that takes into
account the interactions between the control system and the
underlying design of the physical system A comprehensive
review of CCD and handling uncertainties in the formulation
presented in 16 and 17 respectively In this paper we
aim to develop a CCD approach that is suitably used for
BESS design for the OWF Specifically we are interested
in developing CCD approach to handle the challenge of
large system CCD with tight coupling amongst the sub
systems ie MMCs etc We aim to develop a generalized
framework to pose a codesign optimization problem that
can also handle subsystem level coupling constraints during
the control and design optimization Furthermore many
energy systems design requires the system to satisfy multiple
objectives eg low operation cost minimize power loss
etc Few studies on codesign control CCD considering
multiple objectives have been conducted without identifying
the Pareto front 1820 Inspired by recent research on
optimizing marine energy kites 21 we propose a CCD
approach that addresses multiple objectives to find Pareto
solutions for integrating offshore wind farms into the grid
Therefore in this paper we aim to address the aforemen
tioned research objective by developing the CCD approach
using the graphbased optimization framework which can
allow scalability of optimization in the case of large systems
and allow formulating a subsystem level codesign problemarXiv240610365v1 eessSY 14 Jun 2024Fig 1 A schematic for the graphbased optimization formulation is shown
Each node defines a subsystem with individual objective functions
Further we also develop a gradientbased approach to handle
multiple objectives for Pareto set identification
The rest of the paper is organized as follows In SectionII
we discuss the proposed methodology of using graphbased
optimization and the gradient based multiobjective optimiza
tion approach for Pareto set identification In SectionIII we
describe the OWF use case for doing CCD and developed
optimization formulation In SectionIV we discuss the
results from the approach and present our conclusions and
future work in SectionV
II M ETHODOLOGY
We describe the details on the proposed methodology
developed for doing CCD OWF interconnected with ACgrid
with BESS
A GraphBased Optimization
Graphbased modeling abstractions have recently been
explored in convex optimization 22 infrastructure networks
23 supply chain planning problems 24 and simulation
of partial differential equations 25 These abstractions
structures are directly tied to physical topology of the sys
tems Recent work by 23 26 27 has shown that the
optimization and simulation for complex systems can be
represented using the graphbased computational framework
It provides a coherent strategy to capture the modeling
elements for a system which are often common in most
engineering applications Figure 1 shows the graphbased
representation of optimization problem consisting of a set
of nodes and edges Each node represents an individual
subsystem optimization model with variables objectives
constraints and data and each edge captures connectivity
between node models and coupling constraints Once the
graph is constructed it can be communicated to traditional or
decomposition optimization solvers eg Groubi orIpopt
In CCD problem for OWF connecting to AC grid through
MTDC we use the system topology of MTDC and AC grids
to define the nodes and edges Each node corresponds to a
subsystem level representation The details about the CCD
formulation and how to convert it to a graphbased model
will be discussed in sectionIII
Next we describe the details on the proposed gradient
based approach for solving multiobjective optimizationB Gradient based Multiobjective optimization
We propose a gradientbased multiobjective optimization
framework inspired by the weightedsum method and the
bilevel optimization algorithm in 28 Given Nobjective
functions that can be split into nnodes the weightedsum
combination of single objective functions f1 f2 f n
g1 g nh1 h ncan be written as
min w1f1f2 fn w2g1g2 gn
wNh1h2 hn 1
where fghare groups of individual objective
functions into different nodes Then the proposed multi
objective optimization framework can be applied to update
the weights to find the Pareto frontier as in the following
Algorithm 1
Algorithm 1 Gradientbased Approach for Multi
Objective Optimization
1Set the step sizes βfor updating w
2Solve the graphbased problem with initial weights
w0 w1 w2 w Nusing Plasmojl
3fork0toKby1do
4 Update wwith projected gradient descent
wk1projNwkβhk
w

8 considered the BESS sizing problem for an onshore wind

farm however the authors developed a sequential approach

where they first chose BESS size and then proposed a control

strategy To the best of the authors knowledge none of

the studies has considered simultaneously accounting for the

control operations of offshore wind farms while sizing BESS

Conventionally the design problem is solved first followed

by operation control optimizations However many studies

915 have shown that such a sequential approach re

sults in suboptimal system performance Control codesign

CCD is a control system design approach that takes into

account the interactions between the control system and the

underlying design of the physical system A comprehensive

review of CCD and handling uncertainties in the formulation

presented in 16 and 17 respectively In this paper we

aim to develop a CCD approach that is suitably used for

BESS design for the OWF Specifically we are interested

in developing CCD approach to handle the challenge of

large system CCD with tight coupling amongst the sub

systems ie MMCs etc We aim to develop a generalized

framework to pose a codesign optimization problem that

can also handle subsystem level coupling constraints during

the control and design optimization Furthermore many

energy systems design requires the system to satisfy multiple

objectives eg low operation cost minimize power loss

etc Few studies on codesign control CCD considering

multiple objectives have been conducted without identifying

the Pareto front 1820 Inspired by recent research on

optimizing marine energy kites 21 we propose a CCD

approach that addresses multiple objectives to find Pareto

solutions for integrating offshore wind farms into the grid

Therefore in this paper we aim to address the aforemen

tioned research objective by developing the CCD approach

using the graphbased optimization framework which can

allow scalability of optimization in the case of large systems

and allow formulating a subsystem level codesign problemarXiv240610365v1 eessSY 14 Jun 2024Fig 1 A schematic for the graphbased optimization formulation is shown

Each node defines a subsystem with individual objective functions

Further we also develop a gradientbased approach to handle

multiple objectives for Pareto set identification

The rest of the paper is organized as follows In SectionII

we discuss the proposed methodology of using graphbased

optimization and the gradient based multiobjective optimiza

tion approach for Pareto set identification In SectionIII we

describe the OWF use case for doing CCD and developed

optimization formulation In SectionIV we discuss the

results from the approach and present our conclusions and

future work in SectionV

II M ETHODOLOGY

We describe the details on the proposed methodology

developed for doing CCD OWF interconnected with ACgrid

with BESS

A GraphBased Optimization

Graphbased modeling abstractions have recently been

explored in convex optimization 22 infrastructure networks

23 supply chain planning problems 24 and simulation

of partial differential equations 25 These abstractions

structures are directly tied to physical topology of the sys

tems Recent work by 23 26 27 has shown that the

optimization and simulation for complex systems can be

represented using the graphbased computational framework

It provides a coherent strategy to capture the modeling

elements for a system which are often common in most

engineering applications Figure 1 shows the graphbased

representation of optimization problem consisting of a set

of nodes and edges Each node represents an individual

subsystem optimization model with variables objectives

constraints and data and each edge captures connectivity

between node models and coupling constraints Once the

graph is constructed it can be communicated to traditional or

decomposition optimization solvers eg Groubi orIpopt

In CCD problem for OWF connecting to AC grid through

MTDC we use the system topology of MTDC and AC grids

to define the nodes and edges Each node corresponds to a

subsystem level representation The details about the CCD

formulation and how to convert it to a graphbased model

will be discussed in sectionIII

Next we describe the details on the proposed gradient

based approach for solving multiobjective optimizationB Gradient based Multiobjective optimization

We propose a gradientbased multiobjective optimization

framework inspired by the weightedsum method and the

bilevel optimization algorithm in 28 Given Nobjective

functions that can be split into nnodes the weightedsum

combination of single objective functions f1 f2 f n

g1 g nh1 h ncan be written as

min w1f1f2 fn w2g1g2 gn

wNh1h2 hn 1

where fghare groups of individual objective

functions into different nodes Then the proposed multi

objective optimization framework can be applied to update

the weights to find the Pareto frontier as in the following

Algorithm 1

Algorithm 1 Gradientbased Approach for Multi

Objective Optimization

1Set the step sizes βfor updating w

2Solve the graphbased problem with initial weights

w0 w1 w2 w Nusing Plasmojl

3fork0toKby1do

4 Update wwith projected gradient descent

wk1projNwkβhk

w