Exploiting Linkage Information and Problem-Specific Knowledge in Evolutionary Distribution Network Expansion Planning

This paper tackles the Distribution Network Expansion Planning (DNEP) problem that has to be solved by distribution network operators to decide which, where, and/or when enhancements to electricity networks should be introduced to satisfy the future power demands. We compare two evolutionary algorithms (EAs) for optimizing expansion plans: the classic genetic algorithm (GA) with uniform crossover and the Gene-pool Optimal Mixing Evolutionary Algorithm (GOMEA) that learns and exploits linkage information between problem variables. We study the impact of incorporating different levels of problem-specific knowledge in the variation operators as well as two constraint-handling techniques: constraint domination and repair mechanisms. Experiments show that the use of problem-specific variation operators is far more important for the classic GA to find high-quality solutions to the DNEP problem. GOMEA is found to have far more robust performance even when an out-of-box variant is used that doesn't exploit problem-specific knowledge. Based on experiments, we suggest that when selecting optimization algorithms for real-world applications like DNEP, EAs that have the ability to model and exploit problem structures, such as GOMEAs and estimation-of-distribution algorithms, should be given priority, especially when problem-specific knowledge is not straightforward to exploit, e.g. in the case of black-box optimization.