Application of artificial neural networks for management of energy systems with high share of renewable energy sources : master's thesis

Energetske mikromreže imaju fleksibilan sastav, kompleksne mehanizme aktivnosti i veliku količinu podataka tokom rada. Stoga, zahtijevaju složenu analizu kako bi se optimirao rad i akumuliralo znanje o vođenju. Simuliran je energetski sustav od dvije tisuće kućanstava sa visokim udjelom OIE, te na temelju rezultata simulacije bilance energetskih tokova za mikromrežu, konstruirana je umjetna neuronska mreža. Za odabrane ulazne vrijednosti, zadatak mreže je prognozirati buduće vrijednosti kontrolnih varijabli te na taj način demonstrirati sposobnost vođenja, odnosno regulacije. Stvoren je generalni model koji je na temelju analize rezultata daljnje optimiziran te je naposlijetku provedena analiza osjetljivosti za dva tipična parametra. Konstruirani model pokazuje sposobnost učenja i rješavanja zadanog problema sa visokim stupnjem točnosti.Energy microgrids have flexible composition, complex mechanisms of activity, and a large amount of data during operation. Therefore, they require complex analysis in order to optimize operation and accumulate scheduling knowledge. Energy system of two thousand households with a high share of RES was simulated, and based on the results of the simulation of the energy flow balance for the microgrid, an artificial neural network was constructed. For the selected input values, the task of the network is to forecast the future values of control variables and thus demonstrate the ability to control and regulate the microgrid. A general model was created, which was further optimized based on the analysis of the results, and finally a sensitivity analysis was performed for two typical parameters. The constructed model shows the ability to learn and solve a given problem with high degree of accuracy