Graph Learning for Distribution Network Design

The design of mini-grid powerline distribution networks is ripe for optimization. When designing powerline distribution networks, we know where the generation plant will be, and we know the locations of all of our customers. In addition to regulations governing the maximum distance between two powerline poles and the maximum distance from the nearest powerline pole to a customer's home or business, we also need to ensure that the maximum voltage drop under peak load conditions is less than a particular threshold, which may involve installing transformers to step the voltage up and down in different sections of the mini-grid. We need to design the least expensive network to successfully connect all of those customers to the generation plant given those constraints. Currently, this process is done entirely by hand, with no good tools for network optimization. Creating such a tool would both be massively beneficial to the industry while also being publishable in a journal.