Abstract:

Accurately predicting the remaining useful life （RUL） of lithium-ion batteries is important for the efficient and safe operation of energy storage systems. For the deficiencies of existing data-driven methods for RUL estimation， which extract aging features non-comprehensively enough and need to predict health state changes before estimating RUL， a RUL estimation method using multi-dimensional and multi-scale features is proposed in this paper to directly estimate the RUL of a battery using constant-current charging voltage segment data. The model maps RUL by extracting aging features of voltage segments using different scale convolution operation after dimensionally transforming the data. The model is validated using publicly available datasets from the University of Oxford， NASA， and the University of Maryland. The validation results show that the model can directly estimate the RUL of the batteries using the voltage segment data without the need of the current historical SOH of the batteries as the training data， which has higher accuracy and universality compared to fixed scale features based on a single dimension.

Key words: lithium-ion battery, remaining useful life, multi-dimensional and multi-scale features, deep learning