SMART MICROGRID ENERGY MANAGEMENT USING DEEP Q-LEARNING METHOD BASED ON HYBRID INVERTER ENERGY EFFICIENCY
Smart microgrid with renewable energy sources is a solution to the increasing consumption of electrical energy which is followed by depletion of fossil energy reserves. Smart microgrids offers energy independence, flexibility and high reliability with energy management systems. The energy management...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/75401 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Smart microgrid with renewable energy sources is a solution to the increasing consumption of electrical energy which is followed by depletion of fossil energy reserves. Smart microgrids offers energy independence, flexibility and high reliability with energy management systems. The energy management system developed in this study is a smart microgrid system installed at the Energy Management Laboratory, Engineering Physics, Bandung Institute of Technology, which is connected to an external power grid, energy storage systems, and energy resources derived from solar energy. This study is an advanced stage of previous microgrid research with an energy management system that was modeled using deep Q-learning by applying the basic concepts of the Markov Decision Process (MDP) which completely describes states, actions, reward functions, and explicit transition probabilities represented by parameters in neural networks. To improve microgrid performance, a grid feed action will be added to the energy management system being modelled. In addition, to avoid component degradation in the microgrid, the efficiency of charging and discharging microgrid energy storage in the hybrid inverter components will also be considered. The simulation results based on the energy management system model in this study show better performance compared to previous research models at MG intelligent Lab. ME, there is an increase in MG performance of 11,54% for battery utilization (BU), 8,95% for renewable fraction (RF), and 14,26% for photovoltaic utilization (PU).
Keywords: microgrid, energy management, hybrid inverter, charge-discharge, grid feed, deep Q-learning |
|---|