Techno-Economic Modeling: Optimising Cost and Revenue

hostitchAI and Robotics

Oct 23, 2013 (3 years and 9 months ago)

77 views

Intelligent Tools for Techno
-
Economic
Modelling

and

Network Design


Tim Glover

Chris Voudouris

Anthony Conway


Edward Tsang

Ali Rais Shaghaghi

Michael Kampouridis

Network Deployment


Given a new country/city


Where should phone/Internet cover be provided?

Deployment Plan Example


Year 1

Year 2

Year 3

No deployment

Cost
vs

Revenue


Cost


Hard optimization
problem


Very technical


Profitability depend
on it!


Revenue


Based on business
model


Commercial
confidential

-300,000
-200,000
-100,000
0
100,000
200,000
300,000
400,000
0
1
2
3
4
5
6
7
8
9
10
Year

Cash Flow

Investment
Revenue
Net
Cost: Fibre Trenching

(Graph Problem)


The network may
include fibres between
exchanges, roads in a
town, or conduit in a
building.


The task is to minimise:


Trenching; and


costs for fibre optic
network deployment.


Confidential Material

Tim Glover(BT)

Ali Rais Shaghaghi(Essex)

Michael Kampouridis (Essex)

Edward Tsang (Essex)

Intelligent Tools for


Fibre Access Network
Design

Tim Glover(BT)

Ali Rais Shaghaghi(Essex)

Edward Tsang(Essex)


BT
NetDesign

is a software platform for assisting with
physical network design
. It is written as a Rich Client
Platform Eclipse application. The main components
are



an
extensible data model describing networks as nodes
and links


a graphical editor for viewing and editing networks


a problem solving package for representing and solving
network design problems


Fibre Access Network Design using
the BT
NetDesign

platform

For

example,

the

network

may

consist

of

fibres

between

exchanges,

roads

in

a

town,

or

conduit

in

a

building
.

The

problem

is

to

minimise

trenching

and

costs

for

fibre

optic

network

deployment
.

1.Fibre Trenching(Graph
Problem)


Guided Local Search

is
a
Metaheuristic


search
method.


Using solution features to improve the local search
algorithm


Considerable improvement in solver algorithm in
regards to execution time and optimised solution when
compared to the existing BT
NetDesign

algorithms
(Simulated Annealing)


Integration to the current BT
NetDesign

platform




Guided local Search for Graph
Problem


In general, an access fibre network consists of a set of
Customers, a set of Distribution Points (DPs) and a set
of Pick

Up Points (exchanges, or PUPs). These points
are located in a network of roads, and possibly open
spaces. The problem is to construct a tree of fibres that
connects each customer to a PUP, either directly, or via
one or more DPs, that minimises the total cost.

2.Access
Fibre Network Design



Different DPs are available of different capacities (
eg

44, 88).


Different customer types may require different numbers of
connections


Different cables are available that bundle together different
numbers of fibres


Different roads may have different costs associated with digging
trenches


Digging a trench across a road costs more than digging along a
pavement.


There is a maximum reach between customers and DPs, and
between DPs and PUPs


Some considerations affecting
cost


Tightly
constrained problem


In some cases finding a
single feasible solution
could
take months


Conventional
search methods
were unable to
solve the
problem



Use of advanced CS methods to have fast and
optimised solutions


In Cases of extremely tightly constrained problems the
CS solver would ensure that at least one solution is found

Intelligence for solving constraint
satisfaction network design problem


Benefit of joint work to date


This work has mainly focused on introducing novel
intelligent problem solving algorithms to the BT
NetDesign

platform.


They have contributed mainly into two areas


Graph Problem


Access
Fibre Network
Design


Concluding Remarks

Intelligent Tools for Techno
-
Economic
Modelling

and

Network Design


Tim Glover (BT)

Michael Kampouridis (Essex)

Ali
Rais

Shaghaghi

(Essex)

Edward Tsang (Essex)

Techno
-
economic
modelling

for
FTTx


Produce a model which


Analyses the technological requirements of the
deployment of an
FTTx

investment


e
.g. number of workers, trenching length, cable length


Analyses the economical requirements of the above
deployment


e
.g

annual cost, annual revenue, cash flow


Purpose of model: to advice on the viability and
profitability of the investment

Model inputs


Area population


Social category


Competition


Budget


Rental tariffs and number of customers


PAYG tariffs and number of customers


Study period

Model outputs


Annual revenue


Annual cost


Cash flow


Net Present Value


Internal Rate of Return

Need for intelligence


While a techno
-
economic model can evaluate different
deployment plans, the number of such plans can be
very large


e.g. if we plan to roll
-
out to 50 cities within the next 5
years, the number of different deployment plans is 5
50


Computationally expensive to evaluate all available
deployment plans


Question: “What is the deployment plan that offers the
highest profit”?

Adding intelligence


Use different heuristics to locate the optimal
deployment plan


Simple Hill Climbing


Steepest Ascent Hill Climbing


Genetic Algorithms

Heat Map
-
Deployment Plan for
London

Improvement of up
to 18% in the NPV
-
equivalent to
millions of pounds
savings

Graphs


-40000000
-30000000
-20000000
-10000000
0
10000000
20000000
30000000
40000000
50000000
60000000
70000000
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Cash Flow

0
10000000
20000000
30000000
40000000
50000000
60000000
70000000
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Annual Revenue

Conclusion


Use of intelligent methods for finding optimal
deployment plans for
FTTx

deployment


Results show that thanks to the methods used, there
has been an increase in the profitability of the
investment


Presented a techno
-
economic tool for evaluation of
such investment