DEVELOPMENT OF ENERGY MANAGEMENT IN SMART MICROGRID SYSTEMS USING DEEP Q-LEARNING METHOD TO INCREASE RENEWABLE FRACTION AND BATTERY UTILIZATION, AS WELL AS REDUCING THE LEVELZED COST OF ELECTRICITY

• Main Author: Mahardika P. Setiawan, Fadhila
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/81494
• Institution: Institut Teknologi Bandung

<p align="justify">Microgrid (MG), as an intelligent electrical energy system entity that realizes the integration of renewable energy sources, is the answer to the increasing need for electrical energy along with the depletion of fossil energy reserves. However, the intermittent nature of renewable energy sources is an obstacle for MG, because it can cause a decline in MG performance which is characterized by not optimal renewable fraction (RF) and battery utilization (BU). This can be overcome by implementing controls on MG components, one of which is the energy storage battery system (SBPE). Reinforcement learning (RL), a branch of machine learning, is one of the promising methods to be applied to energy management control in intelligent MGs. The use of RL algorithms in smart MGs can create agents that can be trained to carry out a task, in this case regulating the charging and discharging of the SBPE for energy management, to fulfill a desired reward function, namely efficient smart MG operation based on its performance parameters. One of the RL methods that will be used in this research is deep Q-learning. A special deep reinforcement learning (DRL) algorithm called DQN is applied to the optimal energy management of MG with uncertainty. The goal is to find the most cost-effective generation schedule from MG by fully utilizing the energy storage system. In practice, the power capacity produced by solar panels often exceeds the power required by the load. However, the system does not receive the full power produced by the solar panels due to limited load power requirements and battery capacity. This is caused by the derating concept applied to PV inverters where there is a reduction in output power depending on condition requirements as an action to protect components from damage by increasing the inverter frequency. Suboptimal use of PV power causes the renewable fraction (RF) value to be low. One way to increase the RF value is to supply power to the public electricity network or what is called grid feed.<p align="justify">