Optimizing cable routes in offshore wind farms Arne Klein Dag - PowerPoint PPT Presentation

Optimizing cable routes in offshore wind farms Arne Klein Dag Haugland Department of Informatics, University of Bergen, Norway Energy lab, January 17th 2017

Offshore wind farm cabling Motivation ◮ High cabling and trenching costs offshore ◮ Often selected manually ◮ “Free” improvements by applying optimization ◮ Some companies (e.g. Statkraft) started using optimization methods ◮ Creating more advanced models, taking into consideration more aspects

Given data ◮ Wind turbine positions ◮ Substation position(s) ◮ Max. energy output of turbines ◮ Obstacles ◮ (Available cable types) ◮ (Cable paths for comparison)

Wind farm data ◮ Turbine and substation position data of offshore wind farms ◮ Barrow ◮ Sheringham Shoal ◮ Walney 1 ◮ Walney 2 Sheringham Shoal Walney 2

Problem properties Basics ◮ Cable capacity ◮ Connectivity ◮ turbines to substations ◮ Non-crossing Possible additions ◮ Branching ◮ Different cable types ◮ Obstacles ◮ Parallel cables ◮ Energy losses

Problem properties Basics ◮ Cable capacity ◮ Connectivity ◮ turbines to substations We want to ◮ Non-crossing ◮ Find optimal cable paths Possible additions ◮ Minimize total cable ◮ Branching length/cost ◮ Different cable types ◮ Satisfy constraints ◮ Obstacles ◮ Parallel cables ◮ Energy losses

Optimization method and solution method ◮ Mathematical model describing the problem ◮ Integer linear programming (ILP) ◮ Linear constraints ◮ Binary decision variable ◮ y ij = 1 means that there is a cable between turbine j and i ◮ Implemented using Python, solved by IBM CPLEX optimization library ◮ Non-crossing constraints ( O ( | N | 4 )) only added if solution violates them

Experimental results - one cable type ◮ Relative improvement from branching below 1% for all test cases ◮ Example Sheringham Shoal with C = 5 ◮ relative improvement 0 . 72% No branching Branching

Experimental results - two cable type (1) ◮ Cable capacity C > Q , cable cost c ij = 1 . 7 q ij Walney 1 14 C = 5 12 rel. difference [%] C = 6 10 C = 7 8 6 4 2 0 2 3 4 5 6 Walney 2 9 8 C = 5 rel. difference [%] 7 C = 6 6 C = 7 5 4 3 2 1 0 2 3 4 5 6 Q

Experimental results - two cable type (2) ◮ Walney 1, C = 7, Q = 2 No branching Branching

Parallel cables 5 5 5 4 4 4 3 3 3 2 2 2 1 1 1 0 0 0 3 2 1 0 1 2 3 3 2 1 0 1 2 3 3 2 1 0 1 2 3 ◮ Can improve solutions in some special cases ◮ Same mechanism in model allows to handle obstacles better

Parallel cables example, Walney 1 7000 +5.985e6 6000 5000 4000 3000 2000 1000 0 463000 464000 465000 466000 467000 468000 469000 470000 471000

Challenges ◮ Does not scale well with number of nodes ◮ High computational costs ◮ Information on cable cost hard to obtain

Thank you!