Further Enquiries

School of Computer Science
Ingkarni Wardli Building
The University of Adelaide
SA 5005
AUSTRALIA
Email

Telephone: +61 8 8313 5586
Facsimile: +61 8 8313 4366

Renewable Energy

Coordinator: Dr Markus Wagner

Renewable energy is playing an increasing role in the supply of energy worldwide and will help mitigate climate change. We as computer scientists contribute to the mitigation with the development of algorithmic solutions to a wide range of challenges. For example, we have extensive experience in the design optimisation of wind farms and wave energy farms (see below), and in the development of human-readable prediction models.

Carnegie Wave CETO 5 and 6In 2015, we have begun to collaborate with Prof. Ben Cazzolato's Wave Energy Converter Group at the School of Mechanical Engineering, The University of Adelaide. This research was seed-funded by the DVCR's Interdisciplinary Research Fund with $27,000. So far, our studies provide first insights into the layout problem of submerged wave energy converters. It is also the first time multiple three tether buoys are being investigated. All publications are listed at the very bottom of this page.

In order to promote the research on Computational Intelligence methods for their application to the energy production and consumption domain, members of the Optimisation and Logistics Group co-founded the IEEE CIS Task Force on Computational Intelligence in the Energy Domain. Currently, Dr. Markus Wagner is co-chair of this task force. For more information on the events and special issues that this task force organises, please visit the official website.

Layout Optimization of Large Wind Farms

In order to further increase the productivity of wind farms, we need to exploit methods that help to optimise their performance.


In case of problems with the video stream, watch it on Youtube.

We are using a ‘selection of the fittest’ step-by-step approach called ‘evolutionary algorithms’ to optimise wind turbine placement. This takes into account wake effects, the minimum amount of land needed, wind factors and the complex aerodynamics of wind turbines.

The question of exactly where wind turbines should be placed to gain maximum efficiency is highly complex. An evolutionary algorithm is a mathematical process where potential solutions keep being improved a step at a time until the optimum is reached. Metaphorically speaking, it is like parents producing a number of offspring, each with differing characteristics. As with evolution, each population or ‘set of solutions’ from a new generation should get better. These solutions can be evaluated in parallel to speed up the computation.

Using the same principles, we can also solve human problems through computer algorithms. Current approaches to solving this placement optimisation can only deal with a small number of turbines, but we have recently demonstrated an accurate and efficient algorithm for as many as 1000 turbines.

Together with researchers at the Massachusetts Institute of Technology, we have been working on wind turbine placement optimisation, and we are now looking to fine-tune the algorithms even further using different models of wake effect and complex aerodynamic factors.

Demonstration of our Turbine Displacement Algorithm
Demonstration of our Turbine Displacement Algorithm and its infeasible area modelling capability.
The image on the left portrays a satellite image of the Woolnorth wind farm in Tasmania, Australia (© 2011 Google). The right images are examples of the the loose adaptation used in the modelling tool, and from left to right model the scenario at 0 (16.9 MW), 5,000 (17.6 MW), and 20,000 (17.7 MW) evaluations. As the considered wind is predominantly from the western direction (between 120° to 225°), the turbines tend to form in staggered north/south columns while leaving space along the east/west directions.

Software

For the assessment of wind farm layouts, we implemented the Park wake model as presented in "A. Kusiak and Z. Song: 'Design of Wind Farm layout for Maximum Wind Energy Capture,' Renewable Energy, Vol. 35, No. 3, 2010, pp. 685-694." Our cross-validated implementation is available for download as a Netbeans Java project. In addition to the assessment function, the code contains flexible initializers for different layouts, and correctors for infeasible layouts. (Author: Markus Wagner)

We provide the single-objective code from our Renewable Energy Journal 2013 article here.

We also provide the multi-objective code from our GECCO 2013 article here. The code is based on the popular framework jMetal, and we packaged it with our fitness functions (energy production, minimum spanning tree, convex hull) and our Matlab scripts to generate the plots.

Important note: due to an error in equation 18 of the Kusiak-Song-article, their predictions are exacly 15 times too high. Consequently, our code's predictions are exactly 15 times too high as well.

Constraints

In our scenarios, we have the following constraints placed on our optimization function. The first one enforces an upper bound on the area of the farm. This constraint ensures that we can only place a turbine i within a certain area, which is a realistic constraint for most layout problems. For a rectangular farm with length l and width w this constraint is satisfied 0 ≤xi ≤l and 0 ≤yi ≤w, for all turbines i. The second constraint regulates the spatial proximity, as it dictates the minimal distance within which two turbines can be set up. It is satisfied iff (xi−xj)2+(yi−yj)2 ≥ MR (for all i,j), where R is the rotor radius and M is a proximity factor usually decided ahead of the optimization based on the make and model of the turbines used. We use M = 8 based on the industry standard.

In addition to the above constraints, we assume that all turbines have the same power curves (approximated as piecewise linear functions) and that the same wind resource spans the entire farm. To increase accuracy, these resources can be estimated for different parts in the farm. The assumptions can be revised in a very straight forward manner to generate more realistic scenarios.

Usage in Classrooms

We have used this problem (and this code) as a group project in our course on "Evolutionary Computation". For the results, please click here and here.

Publications

  • Fast and Effective Multi-Objective Optimisation of Submerged Wave Energy Converters
    Didac Rodriguez Arbones, Boyin Ding, Nataliia Sergiienko, and Markus Wagner
    Proceedings, Parallel Problem Solving from Nature (PPSN 2016)
    PDF
  • Fast and effective optimisation of arrays of submerged wave energy converters
    Junhua Wu, Slava Shekh, Nataliia Sergiienko, Benjamin Cazzolato, Boyin Ding, Frank Neumann, and Markus Wagner
    Proceedings, Genetic and Evolutionary Computation Conference (GECCO 2016)
    PDF | Matlab code
  • Constrained Evolutionary Wind Turbine Placement with Penalty Functions
    Daniel Lückehe, Oliver Kramer and Markus Wagner
    Proceedings, World Congress on Computational Intelligence: Congress on Evolutionary Computation (WCCI/CEC 2016)
    PDF
  • Frequency domain model of the three-tether WECs array
    Nataliia Sergiienko, Benjamin Cazzolato, Boyin Ding, and Maziar Arjomandi
    Internal technical report (2016)
  • Renewable Energy 2016 Special Issue "Optimization Methods in Renewable Energy Systems Design"
    Paul Kaufmann, Oliver Kramer, Frank Neumann, and Markus Wagner (Managing Guest Editor)
    ScienceDirect | Our special issue received 135 submissions.
  • On Evolutionary Approaches to Wind Turbine Placement with Geo-Constraints
    Daniel Lückehe, Markus Wagner, and Oliver Kramer
    Proceedings, Genetic and Evolutionary Computation Conference (GECCO 2015)
  • Fast and Effective Multi-Objective Optimisation of Wind Turbine Placement
    Raymond Tran, Junhua Wu, Christopher Denison, Thomas Ackling, Markus Wagner, and Frank Neumann
    Proceedings, Genetic and Evolutionary Computation Conference (GECCO 2013)
    PDF | complete jMetal source code with fitness functions and Matlab scripts
  • A Fast and Effective Local Search Algorithm for Optimizing the Placement of Wind Turbines
    Markus Wagner, Jareth Day, and Frank Neumann
    Elsevier: Renewable Energy | arXiv.org | Code: Turbine Distribution Algorithm (TDA)
  • Optimizing Energy Output and Layout Costs for Large Wind Farms using Particle Swarm Optimization
    Kalyan Veeramachaneni, Markus Wagner, Una-May O'Reilly and Frank Neumann
    Proceedings, IEEE World Congress on Computational Intelligence: Congress on Evolutionary Computation (WCCI/CEC 2012)
    PDF | Abstract
  • Predicting the Energy Output of Wind Farms Based on Weather Data: Important Variables and their Correlation
    Katya Vladislavleva, Tobias Friedrich, Frank Neumann, and Markus Wagner
    Elsevier: Renewable Energy | arXiv.org
  • Optimizing the Layout of 1000 Wind Turbines
    Markus Wagner, Kalyan Veeramachaneni, Frank Neumann, and Una-May O'Reilly
    Proceedings, European Wind Energy Association 2011 (EWEA 2011)
    PDF | Poster | BibTeX | Abstract