This paper introduces a novel method for generating synthetic data tailored for distributed energy storage systems. This innovative framework integrates Generative Adversarial Network (GAN) with the Mamba architecture, aiming to address critical challenges such as restricted data accessibility and limited dataset availability in practical deployment scenarios. The proposed framework has two key components: a GAN-based synthetic data generation module and a Mamba charging state estimator. These components work together to produce high-quality synthetic data. An advanced Mamba-GAN architecture trains the generation module, optimizing data fidelity using a min-max adversarial loss function. For Remaining Useful Life (RUL) estimation, a deep Mamba model with a Selective State Space (SSM) layer is developed. This model efficiently analyzes time-series data and performs well in RUL prediction tasks. Experimental results show that the generated data distribution closely matches real datasets. The model’s RMSE is 3.28%, while its MAE is 1.19%, enabling realistic operational data synthesis for distributed energy storage systems. These synthetic datasets support equipment operation prediction, maintenance alerts, and related applications. They help ensure stable equipment operation while extending its lifespan. Future research will focus on multi-modal data fusion techniques and improving the interpretability of generated data.