On spanning-tree recombination in evolutionary large-scale network problems - application to electrical distribution planning

We report the key algorithms involved in the recombination-based evolutionary software developed for planning electrical distribution networks. We focus on the dimensionality problem of large-scale networks and on the specificities of its search space. We report the difficulties in handling topology constraints and present both the geno-type and the operators to overcome such difficulties. The operators are designed to process meaningful topological information as geno-type substructures and turn the radiality and connectivity into genetic transmissible properties. First, a theoretical example is presented to illustrate important differences between other common approaches and the approach taken. Then, a real electrical industry application is presented to illustrate the ability of the approach to handle large-scale distribution-network problems.