GAs are generally useful when a solution is made up of many values, no matter of which type. The reason this optimization technique is described in more detail here is that GAs work for an extremely wide range of problems. For ecological models, this might be much more important than the speed of calculations or the complexity of the programming task, because usually they are first designed to handle simple questions, where strong assumptions made for reasons of transparency lead to nice properties of the search space allowing simpler and faster algorithms. But as successful models are often subject to further development, the optimization problem might change in an unpredictable way, most often losing qualities like linearity, completeness of the search space or changing from one objective to multiple ones. Thus, the robustness of GAs to work for a wide range of problems especially with extremely big search spaces might be more valuable as speed or even quality ofthe solution, particularly in view of huge uncertainties in the data.
To get an idea when a GA might be a good choice, it is useful to think about how nature uses it in the vast space of all possible genomes. The main aspect here is that a sequence of data that occurs in several good solutions is likely to appear also in the global optimum.
For MERLIN, to every single technology exactly one measure's implementation degree can be increased to its maximum value, which is only restricted by the sum of the other measures' implementation degrees and the share of technology, to which the measure can be applied. Therefore, integer values are most useful to describe, which measure shall be used to a bigger extent than in the reference scenario.
Furthermore, the order of the values the algorithm deals with is of importance, because the mating operator is more likely to combine good parts ofdifferent solutions, ifthe information belonging to these parts is positioned in i=i close proximity. Although there are ways to make this problem of advantageous positioning part of the optimization itself, it is rather obvious that technologies with high potential of emission reduction of the same pollutant in the same country will be sensible parts to keep together.
Was this article helpful?