Lithium-ion batteries are used in various products ranging from consumer electronics such as laptops to larger appliances such as battery energy storage systems (BESS). In Hong Kong, BESS has been introduced to power tower cranes in the construction sites and the air conditioning system in the airport. Therefore, accurate battery life prediction is critical to reducing maintenance frequency and cost. This thesis collects battery data through a battery cycle test. The batteries' sizes are different (18650 and 21700), and the chemistries (NMC and NCA) were collected from five lithium battery manufacturers. Divide the cells of each manufacturer into two groups, each with three battery cells, and test them under different conditions, charge in constant current-constant voltage (CC-CV) mode,...[ Read more ]

Lithium-ion batteries are used in various products ranging from consumer electronics such as laptops to larger appliances such as battery energy storage systems (BESS). In Hong Kong, BESS has been introduced to power tower cranes in the construction sites and the air conditioning system in the airport. Therefore, accurate battery life prediction is critical to reducing maintenance frequency and cost. This thesis collects battery data through a battery cycle test. The batteries' sizes are different (18650 and 21700), and the chemistries (NMC and NCA) were collected from five lithium battery manufacturers. Divide the cells of each manufacturer into two groups, each with three battery cells, and test them under different conditions, charge in constant current-constant voltage (CC-CV) mode, and discharge in constant current (CC) mode. This paper uses a neural network and linear regression models to build a predictive model. The battery data of the training set comes from a single manufacturer to make a prediction model, while the verification set is added to the other manufacturers' data. The test results show that the model established by using the neural network has good compatibility. Regardless of whether it is the same manufacturer or chemical composition, the neural network model has a good prediction effect. It also provides a new direction for the next generation of battery prediction models.