Improving Lithium-Ion Battery Management Systems Using Equivalent Circuit Models, Cloud Platforms, and Machine Learning Estimation Techniques

Abstract

Building a future that preserves the environment and reduces dependence on fossil fuels is an imperative undertaking, and it greatly hinges on the global transition to renewable energy. Energy storage plays an important role in the adoption of renewable energy to help solve climate change problems. Lithium-ion (Li-ion) batteries are an excellent solution for energy storage due to their properties including high energy density, high power density, long cycle life, low self-discharge rate, no memory effects, and low environmental pollution. In order to ensure the safety and efficient operation of Li-ion battery systems, battery management systems (BMSs) are required. However, the current design and functionality of the BMS suffer from a few critical drawbacks including its low computational capability and limited data storage. The work in this thesis focuses on improving the BMS by investigating and improving the equivalent circuit model (ECM) which is often the core battery model used in practical BMS, researching and developing a smart BMS utilizing the cloud platform, and proposing potential applications of the cloud-based smart BMS such as cell replacement or SOH estimation using machine learning. One of the main focuses of this work is on the critical role of accurate battery modeling for safe and effective operation. The first contribution of this research is an investigation into the performance of three ECMs—1RC, 2RC, and 1RC with hysteresis—across four common Li-ion chemistries: LFP, NMC, LMO, and NCA. Experimental results demonstrate that all three models can be applied to these chemistries with low error rates, particularly under dynamic current profiles. The findings indicate that the 1RC with hysteresis ECM performs best for LFP and NCA chemistries, while the 1RC ECM is most suitable for NMC and LMO chemistries. These insights are crucial for optimizing BMS applications in real-world scenarios, highlighting the need to tailor ECM selection to specific battery chemistries. This work also seeks to add further to the improvement of the ECM used in the BMS, as another novel contribution. Its next research delves into the effects of state of charge (SOC), temperature, and state of health (SOH) on the parameters of the ECM, particularly the Thevenin model. While SOC and temperature are well-integrated into ECMs, the impact of SOH has been less explored. Through a series of experiments, it was found that as SOH decreases, both the ohmic and polarization resistances increase, while the polarization capacitance decreases. An empirical model was developed to represent the combined effects of SOH, SOC, and temperature on ECM parameters. This model was validated experimentally and showed significant improvements in accuracy without increasing complexity. The proposed model offers a practical solution for real-world BMS applications, enhancing the precision of battery monitoring and control. As part of the next steps in the research, the concept and development of cloud-based smart BMSs were discussed in detail. Traditional BMSs are limited by computational capacity and data storage, which can hinder the development of advanced battery management algorithms. A cloud-based BMS can address these issues by offloading computation and storage to the cloud, enabling more sophisticated and reliable battery algorithms. The study discusses the design, functionality, and benefits of cloud-based BMSs, including improved reliability and performance of Li-ion battery systems. It also explores the division of tasks between local and cloud functions, emphasizing the potential for significant advancements in battery management through cloud integration. This innovation is expected to play a pivotal role in advancing renewable energy technologies. Once the development of the cloud-based smart BMS has been discussed, the practical applications of such innovation are then examined. As the optimization of Li-ion battery pack usage becomes increasingly more necessary and given the inevitable degradation of Li-ion batteries, recent research has focused on maximizing the utilization of battery cells within packs. Another study, which is the next novel contribution of this work, investigates the feasibility and benefits of cell replacement within battery packs, using a simulation framework based on cell voltage and degradation models. The cloud-based BMS, with more data storage, shall be able to advance the cell replacement application by providing significantly more battery historical data. The study, conducted using MATLAB, simulates the life cycles of battery packs with varied cell configurations. Results show that cell replacement can significantly extend the lifespan of battery packs and is economically advantageous compared to a full pack replacement. For practical implementation, the design criteria include individual cell monitoring and easy accessibility for cell replacement, underscoring the potential for more efficient battery usage strategies. Another practical application of the cloud-based BMS with improved computational and memory capability is battery state estimation, specifically complex algorithms such as SOH estimation. A part of this work, its final novel contribution, develops a novel SOH estimation approach, which is crucial for the effective operation of Li-ion battery systems. Existing methods have limitations in adaptability and real-time application. This study introduces a machine learning-based approach for online SOH estimation during fast charging cycles. Using a dataset of 124 cells, with various machine learning algorithms tested, the neural network algorithm demonstrated superior accuracy with an RMSE of 9.50mAh and a MAPE of 0.69%. The methodology, which utilizes partial charge metrics without needing historical data, is highly suitable for real-time BMS integration. This approach enhances the reliability and performance of Li-ion battery systems, contributing to the broader adoption of electric vehicles (EVs) and renewable energy technologies. Overall, this thesis presents significant advancements in the field of battery management systems, particularly through the improvement of the ECM, the introduction of cloud-based smart BMSs, and the development of innovative cell replacement and SOH estimation methods. The cloud-based BMS would be able to solve the problems of limited computational capability and data storage. It would also lead to more accurate and reliable battery algorithms and allow the development of other complex BMS functions. The cloud-based smart BMS is expected to improve the reliability and overall performance of Li-ion battery systems, contributing to the mass adoption of EVs and renewable energy.