State of Health Estimation of Microgrid Energy Storage Batteries Based on An Improved Temporal Convolutional Network

The energy storage batteries are important part of the power grid and microgrid. Estimating the state of health (SoH) is a key technology for ensuring the safety and efficiency of battery power supply. In this work, we propose an improved temporal convolutional network (TCN) for SoH estimation. Firstly, we extract aging features from the discharge curves of the batteries to reflect their health status. Specifically, the discharge capacity, discharge voltage and discharge time are recorded and chosen as features at regular intervals. Next, we present a TCN -based approach for SoH estimation, incorporating an attention mechanism to enhance the TCN structure and improve estimation accuracy. Finally, the TCN-attention model was trained from 28 batteries, while 8 batteries are utilized to validate the model's performance. The results indicate that the proposed approach exhibits excellent state of health estimation ability compared to other commonly-used algorithms and the TCN method without attention mechanism.