During my senior undergraduate year at OSU, I worked with the COST Lab under the advisement of Dr. Stockar and Audrey Blizard developing genetic algorithms to solve mathematical partitioning problems. This research was one supporting element of Ms. Blizard's dissertation work, which was focused on optimizing the control of distributed heating networks using a partition based approach. See their associated publication for additional information on this problem and the techniques used to solve it.
My contribution was in helping select an optimal set of partitions for a given distributed heating network. I used a genetic algorithm approach to solve this problem. The algorithm begins with a set of nodes grouped repeatedly into many different random partitions, called partition sets. These parent sets are used to create several child sets via mutations including jumps (moving a node to a different partition) and crossovers (swapping the partitions of two nodes simultaneously). Each child set is then evaluated using a cost function, and a group of the highest performing sets are used as the parent sets for a new round of mutations. The initial sets only use two distinct partitions, but sets can additionally be mutated by creating "subsets", i.e. splitting a partition into two smaller partitions. The best performing partition sets for each number of partitions are saved separately; different partition counts may be more or less optimal for different use cases.
This algorithm was implemented in MatLab using CasADi to help optimize the generated partition sets. Unfortunately, the code for this implementation is not public. If you are a potential employer or collaborator, please reach out to me privately, so that I can work with COST to provide you additional information on this project.