240 likes | 353 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.
E N D
Intelligent Tools for Techno-Economic ModellingandNetwork Design Tim GloverChris 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
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 forFibre Access Network Design Tim Glover(BT) Ali Rais Shaghaghi(Essex) Edward Tsang(Essex)
Fibre Access Network Design using the BT NetDesignplatform • BT NetDesignis 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
1.Fibre Trenching(Graph Problem) 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.
Guided local Search for 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
2.Access Fibre Network Design • 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.
Some considerations affecting cost • 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
Intelligence for solving constraint satisfaction network design problem • 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
Concluding Remarks • 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
Intelligent Tools for Techno-Economic ModellingandNetwork Design Tim Glover (BT)Michael Kampouridis (Essex)Ali RaisShaghaghi (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 550 • 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
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