ANOMALY DETECTION AND FAILURE POTENTIAL ANALYSIS OF BATTERY ENERGY STORAGE SYSTEM (BESS) USING FAILURE MODE AND EFFECT ANALYSIS (FMEA) METHOD

• Main Author: Intan Patya, Dhea
• Format: Theses
• Language: Indonesia
• Subjects: Teknik (Rekayasa, enjinering dan kegiatan berkaitan)
• Online Access: https://digilib.itb.ac.id/gdl/view/80530
• Institution: Institut Teknologi Bandung

Electrical energy storage systems can be done using batteries. Batteries are a source of chemical-electrical energy that can store energy and convert it into electric power. In the development of energy demand, batteries have a very important role, such as in electric vehicles and solar power plants. The battery that is often used is the lithium-ion battery type. Good performances battery, will support its the devices. Lithium-ion batteries are referred to as batteries that have a long life and several other advantages such as being environmentally friendly, having a large amount of power and energy and quite guaranteed safety, but there are several factors that can cause problems and reduce performance, or result in failure. what's wrong with the battery. The amount of energy that can be stored in a battery is limited, so the battery will experience a cycle of charge and discharge. A series of activities on the battery such as overcharge, over-discharge, and high temperature differences often occur in the battery. The process of activity on the battery that is not right can cause battery performance to decrease or the occurrence of failure of the battery and cause anomalies in its voltage, current, and temperature. This method of diagnosing battery system failures can contribute to a more optimal battery maintenance schedule, so that performance is maintained. Anomaly detection method based on FMEA to detect failure can be in the form of boundary conditions of battery protection (fixed value of range-of-change) from battery performance parameters. The application of the anomaly detection method using FMEA in this study is expected to be able to monitor if there is a decrease in performance in the SBPE from the battery system so that potential failures can be avoided as early as possible. However, because the nature of FMEA is too conceptual and can only be done manually, machine learning methods are needed to obtain FMEA documents automatically. The supervised learning type of machine learning method is used because it is more accurate and faster in the case of predicting input to output achieved. Machine learning methods used include K-Nearest Neighbor (KNN), Random Forest (RF) and Supporting Vector Classification (KVP). The research flow was carried out using the Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology. In the business understanding, it is explained about monitoring work on battery cells and detecting anomalies in measuring temperature values in batteries based on a statistical approach to calculating standard deviations. SBPE consists of 18 modules where each module consists of 15 battery cells, so there are 270 LiFePO4 3.2 VDC 100Ah battery cells in 1 cluster. The understanding of the data describes the type, structure and quality of the data used. Next, the labeling process is carried out at the data preparation stage. Then modeled using machine learning, namely KNN, RF and KVP to obtain predictions of SEV, OCC, DET, and RPN values. The input target used is temperature. The best model obtained is a model using the KNN method. After that, an evaluation and analysis of anomaly detection is carried out using FMEA analysis by calculating the values of severity, occurrence, and detection which will then produce a Risk Priority Number (RPN) value by sorting the system or component that has the highest failure rate or anomaly. In addition, a statistical approach is used to determine the outlier data limit on the SBPE temperature measurement value. Values from the calculation results that exceed the values of two and three standard deviations are classified into battery status labels, with labels 1 (normal), 2 (warning), and 3 (anomaly). Based on the results of machine learning methods and statistical approaches, the highest level of risk of potential failure is in SBPE cluster 2, cabinet 4, modules 2 and 3. Potential failures that occur in cabinet 4 modules 2 and 3 are the failure of the temperature sensor, imbalance on battery cell charges due to unequal battery conditions, as well as the potential for battery bloating due to heat release events and gas reactions in the battery.