PERFORMANCE PREDICTION OF RENEWABLE FRACTION FOR SMART ELECTRICITY GRID USING HYBRID MODELLING METHOD

• Main Author: Narendra Purnama Putra, Taufiq
• Format: Final Project
• Language: Indonesia
• Subjects: Teknik (Rekayasa, enjinering dan kegiatan berkaitan)
• Online Access: https://digilib.itb.ac.id/gdl/view/79274
• Institution: Institut Teknologi Bandung

Indonesia National Energy Planning requires to achieve the New and Renewable Energy (NRE) fraction for electricity generation by 23 % in 2025. The intermittent nature of NRE can be overcome by implementing energy management algorithm to smart grid. To evaluate the target, Renewable Fraction (RF) will be used and predicted in this research. This research will be conducted using Cross Reference Industry Standard Process for Data Mining (CRISP-DM) methodology. The methodological steps are as follows : (1) determining the purpose of this research (business understanding), (2) analyzing the system’s components and operational data they generate (data understanding), (3) preparing the raw operational data to be able to be used for modeling (data preparation), (4) modelling the system using hybrid modeling (modeling), (5) evaluating the model (evaluation), and (6) calculating the RF based on the modeling results using two methods, namely via SoC and direct (deployment). To model the RF, combinations of machine learning (ML) and physics-based model (model based, MB) are used. There are three main components that need to be modeled, which is Photovoltaic (PV), Battery Energy Storage System (BESS), and electricity load. BESS can be modeled using two reference, cell battery data and the BESS as a whole. Electricity load and BESS as a whole are modeled using hybrid modeling. Where as BESS modeling with the reference of cell battery data is modeled using ML. PV is modeled using MB (PVLib). Overall, it can be concluded that the combinations of the usage of ML and MB to model the RF implies that this research is carried out using the hybrid modeling method. Results yielded from this research is that the average RF value on 7th - 13th of January 2018 is 9.28 % (direct) and 4.62 % (via SoC). When projected to the year of 2020 and taking account the degradation of battery and solar cells in the system, the RF value decreased for 6.08 % (direct) and 1.41 % (via SoC). Therefore, the target in the Indonesia National Energy Planning have not been met in this research.