The assessment of state of health (SOH) of lithium-ion batteries (LIBs) is crucial for ensuring their safety and efficiency. In this study, we an enhanced data-driven approach for time-series SOH prediction. The time-series battery data is then improved by applying the bidirectional long short-term memory positional encoding attention (BiLSTM-PEA). This transformation makes it possible to process the data more effectively and greatly improves the model's capacity to capture intricate temporal dynamics and dependencies. By offering vital contextual information regarding the relative position and order of each piece in the dataset, it enhances the input. The weight of each input variable is then dynamically changed at each discrete time by the model's unique embedding of a time-step internal attention mechanism to this enhanced data. In the end, this improved input is processed by the BiLSTM-PEA networks, which capture long-term relationships and enable the successful modeling of complex patterns in the data that may be utilized to forecast the SOH of LIB. © 2024 IEEE.