Status
Principle Investigator Contact Information
Name: Markus Wagner
Address: School of Computer Science, Ingkarni Wardli Building, 4th floor, room 4.45
The University of Adelaide, Australia, Adelaide, SA 5005, Australia
Phone: +61 8 831 35405
Email: markus.wagner@adelaide.edu.au
Description
The southern coast of Australia is broadly identified as one of the best candidates for installing the wave energy farm because of the consistently large waves. Therefore, four real wave scenarios from the southern coast of Australia (Sydney, Adelaide, Perth and Tasmania) are simulated to evaluate the applied Wave Energy Converters (WEC) performances. The employed converter model is a fully submerged three-tether converter called CETO. Three optimization aspects of the CETO have been considered including layout (positions), power take-off (PTO) and geometric parameters (shape). According to the optimization results, the new achieved configurations can produce more energy substantially compared with the previous research.
Funding Source
Deputy Vice-Chancellor (Research) (DVCR)
Location of Research
Australia
Project Aims
1. To create the real wave energy resource model for four sites (Sydney, Adelaide, Perth and Tasmania).
2. To evaluate the performance of WECs by these real wave scenarios and analyzing the power landscape.
3. To maximize the total absorbed power by optimizing the position of the layouts with different numbers of converters.
4. To investigate and propose better power take-off configurations in each real wave model.
5. To optimize the geometric parameters of WECs for developing the harnessed energy by each converter.
Study Progress
Two initial parts of the project are completed and awaiting publications. Furthermore, the third part is started.
Key Findings
WP1. A hybrid method of stochastic local search combined with Nelder-Mead Simplex direct search is proposed for maximizing the total absorbed energy. It performs better than previous search techniques for exploring the search space of 4 and 16 converters interactions. The optimization results show that the hybrid framework is a fast and effective idea for placing WECs in a size-constrained environment.
WP2. According to many observations of the wave farm power landscape, a proper search sector is proposed for exploring and placing the next sequential WEC by a smart local search approach. The method considerably outperforms all previous heuristic methods in terms of both quality of achieved solutions and the convergence-rate of search in all tested wave regimes.
WP3. The absorbed power of the wave farm is modified by optimizing both WECs locations and PTO parameters. A modern hybrid heuristics is suggested that is a combination of a symmetric local search and a numerical optimization idea. The results in this study show that the search problem is challenging, with buoys inducing changes in the local power landscape and hydrodynamic interactions occurring between buoys. The PTO optimization outcomes, also, indicate at least some interaction between buoy position and optimal PTO settings for each buoy. Moreover, the hydrodynamic modelling needed for larger buoy layouts is expensive, which constrains optimization to take place with a limited number of evaluations.
WP4. As simulating and computing the complex hydrodynamic interactions in the wave farms is computationally expensive, which restricts optimization methods to have just a few evaluations. For dealing with this expensive optimization problem, an adaptive neuro-surrogate optimization (ANSO) method with a very limited number of observations. The trained model is applied using a greedy local search with a backtracking optimization strategy. For evaluating the performance of the proposed approach, some of the more popular and successful Evolutionary Algorithms (EAs) are compared in four real wave scenarios (Sydney, Perth, Adelaide and Tasmania). Experimental results explain that the adaptive neuro model is competitive with other optimization methods in terms of total harnessed power output and faster in terms of total computational costs.