Neural network-aided hourly dispatch optimization of a droop-controlled islanded microgrid

• Main Author: Torrizo, Lorwin Felimar B.
• Format: text
• Language: English
• Published: Animo Repository 2020
• Subjects: Microgrids (Smart power grids); Neural networks (Computer science); Electrical and Computer Engineering; Systems and Communications
• Online Access: https://animorepository.dlsu.edu.ph/etd_masteral/5973; https://animorepository.dlsu.edu.ph/context/etd_masteral/article/12919/viewcontent/Torrizo_LorwinFelimarB_11791497_Neural_Network_aided_Hourly_Dispatch_Optimization_of_a_Droop_Controlled_Islanded_Microgrid_1_Redacted.pdf
• Institution: De La Salle University

This study primarily aims to develop a diesel consumption-minimizing optimization algorithm for a droop-controlled islanded microgrid that utilizes an artificial neural network for determining the next-hour dispatch. Multiple activities were done in order to achieve this, including gathering various meteorological data and representative demand profile, constructing the neural network, and defining the optimization problem and its constraints. Three microgrid configurations were used to evaluate the impact of the optimization algorithm. During the process of the study, the standalone load and solar forecasting neural networks were evaluated against the results from their adapted study. The study’s neural networks proceeded to display an improvement against its reference, posting MAPE values of 2.194% and 0.925% respectively. After executing further adjustments to increase the reliability of the dispatch to as much as 99.7%, the study’s optimization derived dispatch utilizing forecasted values posted as much as 11.9% less diesel consumption than the reference dispatch algorithm.