In this paper, a novel Chinese word segmentation (CWS) model, named as the optimized Attention-Bidirectional - Long Short-Term Memory-Conditional Random Field (Attention-BiLSTM-CRF), is specifically designed for substation configuration description (SCD) text data in smart grid station. In the proposed model, the Skip-Gram algorithm is utilized to convert word data into distributed vectors, then by integrating the attention mechanism, LSTM, and CRF, the new segmentation model named as Attention-BiLSTM-CRF is developed, which is more suitable for SCD text data in smart grid station. Finally, the effectiveness of the proposed approach is demonstrated through a case study based on SCD text data. Simulation results show that the proposed model can achieve a higher accuracy rate of 98.2%, and it can be observed that the proposed CWS approach had higher accuracy and speed in the field of smart grid.