LEE- Elektrik Mühendisliği Lisansüstü Programı

Bu topluluk için Kalıcı Uri

http://hdl.handle.net/11527/19259

Gözat

Yazar "Altun, Osman Alper" ile LEE- Elektrik Mühendisliği Lisansüstü Programı'a göz atma
  • Öge
    State of charge estimation of lithium-ion batteries using machine learning approach
    (Graduate School, 2024-02-06) Altun, Osman Alper ; Ayaz, Emine ; 504201045 ; Electrical Engineering
    The issues of global warming, environmental pollution, and the fast depletion of resources caused by the use of fossil fuels have prompted scientists and governments to explore several solutions for these challenges. Specifically, it is crucial to minimize the impact of vehicles used in transportation in this context. Currently, electric vehicle are swiftly replacing internal combustion engine vehicles. Governments provide a range of incentives and rules for electric cars, which effectively decrease carbon emissions, minimize greenhouse gas emissions, and address the issue of noise pollution. As an example, starting from 2030, the United Kingdom will prohibit the sale of new internal combustion vehicles. Collaboration across several sectors is necessary for the advancement of electric vehicle technology. The primary challenge associated with electric cars now lies in the battery, which exhibits variability in performance based on use and environmental factors. Lithium-ion (Li-ion) batteries are the most favored form of battery for Electric Vehicles (EVs). Li-ion batteries are chosen over lead acid and NiMH batteries owing to their minimal self-discharge, high energy density, extended lifetime, and quick charging capabilities. To provide the necessary capacity demanded by vehicles, several Li-ion cells are interconnected in both series and parallel configurations. The reliable and secure functioning of these many cells is crucial for the vehicle's performance and the users' safety. The battery management system (BMS) is responsible for establishing the connection between all controls inside the battery and various peripheral systems in the vehicle. BMS, which has a main objective of ensuring vehicle and driver safety, is a combination of hardware and software components. Stable functioning is essential for optimal vehicle performance. It carries out several functions like cell monitoring, state estimation, cell balancing, thermal management, and charging control, covering from individual cells to the whole battery pack. The main subject of our thesis will be the SOC, which is one of the state estimates we will examine. SOC serves as the battery charge status indicator in EVs, similar to how the fuel gauge functions in internal combustion automobiles. As a result of the characteristics of the battery, the SOC cannot be directly measured, but it can be approximated using several techniques. The techniques for estimating SOC may be divided into two groups: Direct Measurement Methods and Indirect Measurement Methods. Direct techniques for generating predictions use the quantifiable properties of the battery. The metrics in issue might include both current and voltage. However, these approaches with reduced complexity also include some drawbacks. Disregarding the battery's aging is a contributing element that will increase the mistake rate in SOC assessment as the battery's use duration extends. The Open Circuit Voltage (OCV) approach is widely used as one of the primary direct methods. The OCV method assesses the battery's charge level by examining its stable condition after an acceptable duration of time. This method is straightforward and accurate, but time-consuming. The procedure relies on constant monitoring of the cell voltage and using SOC values acquired from tables. The Coulomb Counting technique is another highly favored approach. The fundamental premise of this technique is to compute the total amount of electric current that flows through and out of the battery during a certain time frame. Indirect approaches, as opposed to direct ones, are more complicated yet provide superior outcomes in the estimate of SOC. Predictions in model-based techniques are generated by constructing many battery models. Adaptive approaches are used in conjunction with both direct and model-based procedures. By possessing the capacity to adjust to evolving circumstances, the drawbacks associated with other approaches are eradicated. Utilizing a feedback mechanism greatly enhances the predictive precision in research employing adaptive filter-based systems.