Genetic Algorithm-based Optimization for Battery Scheduling in a Smart Distribution Grid

Sammanfattning: In recent years, with the introduction of renewable energy sources (RES) and digital technologies, grids in Europe have become smarter.This is motivated by the need to effectively use the infrastructure incases of high penetration of RES. Consumer participation is one of the important features of smart grids. This customer participation can be individually controlled or collectively controlled by an aggregator. This thesis simulates a neighbourhood where each house is equipped with a photovoltaic (PV) system and a battery storage system such that each house is a net-zero energy building (NZEB). A comparison is made between the cost savings for the households when the battery is individually controlled and when an aggregator controls all batteries. The battery scheduling is optimized using genetic algorithm(GA). The results show higher savings in the case of aggregator control: the self-consumption (SC) increased with 4 percentage points, and the total energy costs for the whole neighbourhood was reduced by more than 40%. This thesis prepares the training data needed to implement the energy management system (EMS) using recurrent neuralnetworks (RNNs).