CNN-LSTM Based Capacity Eatimation of Lithium-ion Batteries In Charging Profiles

Lithium-ion batteries have attracted widespread interest due to their excellent cycling performance and high energy density. With long-term use, batteries inevitably age, and battery capacity estimation is crucial to the safe operation of batteries. Battery capacity degradation is characterized by capacity regeneration and calendar aging, and most of the existing work targets the calendar aging. This paper takes both characteristics into account, we proposes a CNN-LSTM neural network model for battery capacity estimation is applied to analyze the aging characteristics of batteries. CNN suppresses high volatility of prediction results by spatial transformation of data, and LSTM predicts calendar aging trend of battery capacity. The model is trained and validated using current, voltage, and temperature that can be directly measured in battery management system. The experimental results show that the model can effectively estimate the battery capacity. CNN-LSTM can solve the above problems simultaneously, and the experimental results show that the proposed method has more accurate estimation results by comparing with Conv1D, Conv2D and LSTM methods.