Measurement, Modelling and State Estimation Techniques for Lithium-ion batteries

Publication Type:
Issue Date:
Full metadata record
Lithium-ion batteries have been widely adopted in energy storage systems for electric vehicles (EVs), electric portable devices, smart grid, and renewable energy systems because of their high energy density, long lifetime, and low self-release rate. When lithium-ion batteries are used in real applications such as EVs, they normally work with power converters, which can deliver power from the batteries to the load and regulate the system output voltage. However, lithium-ions batteries also have a critical safety concern. As chemical products, the battery states such as state of charge (SOC) cannot be directly measured by sensors; the only directly measurable signals of lithium-ion batteries during battery operation are terminal voltage, operational current and temperature. Some models have been established to calculate information of the battery states using measured signals. However, the inherent chemical characteristics of lithium-ion batteries mean that it is difficult to achieve a highly accurate online battery state monitoring or estimation. When the battery state is estimated inaccurately, it will waste the available capacities, reduce battery lifetime, and could even lead to fire or explosion. To avoid these issues, lithium-ion batteries should be well- monitored and managed by a battery management system (BMS). This thesis focuses on improving the efficiency, reliability and estimation accuracy for the BMS of lithium-ion batteries from signals measurement, to battery modelling to state estimation perspectives. First, this thesis develops an improved battery modelling techniques by proposing a rapid and accurate open circuit voltage (OCV) measurement method. Second, this thesis develops practical battery impedance measurement techniques, which can be used for offline battery modelling and online states monitoring. Third, a sensorless battery surface temperature estimation has been proposed to improve the reliability and reduce the cost of the BMS. Fourth, as artificial intelligence (AI) technology has developed, more recurrent neural network (RNN) based battery SOC estimation methods have been proposed. This thesis comprehensively evaluates previous methods from theoretical and experimental perspectives and proposes a RNN model with a suitable hyper-parameter setting for online SOC estimation with high accuracy and low computational burden.
Please use this identifier to cite or link to this item: