We were curious as to whether TDA s lower performance

We were curious as to whether TDA\'s lower performance was a consequence of our implementation of it, or a genuine reflection of TDA\'s true likely performance. To that end, we examined the result of the 2014 Wind Farm Layout Optimisation AR-R 17779 in which TDA was an entrant. The results of TDA on the five scenarios, as reported in the competition results, are given in Table 5. They show that, if any conclusion is to be drawn, it is that our implementation of TDA actually slightly outperforms the version used in the competition. Specifically, the median results in the plots are actually slightly higher than the wake free ratios shown in Table 5.
Table 5.
Reported performance of TDA in the 2014 Wind Farm Layout Optimisation competition.ScenarioReported best wake free ratio10.915120.911230.853540.877750.8373Full-size tableTable optionsView in workspaceDownload as CSV
It is interesting to speculate as to the reason why TDA\'s performance is below that of the other algorithms. In our opinion, TDA\'s performance is related to the size of the layouts. Scenario 2 in the evaluation set is the layout with the smallest number of turbines (only 150) and Fig. 5(b), which concerns this scenario, shows that TDA is comparable to the other algorithms. Since 1000 evaluations are performed per run, T cells can be expected that TDA will mutate each individual turbine in Scenario 2 1000/150=6.661000/150=6.66 times on average per run. However for the largest layout (Scenario 5 with 910 turbines), the number of expected mutations per turbine drops to 1000/910=1.101000/910=1.10. TDA therefore may be the optimal choice for smaller layouts, but for larger layouts it suffers because it requires more evaluations to achieve the same degree of position tuning. A future modification to the TDA algorithm could alleviate this problem.