Accurate state of charge (SOC) estimation of lithium-ion batteries is essential for preventing overcharging and over-discharging, thereby ensuring safety and reliability in battery-powered systems. In this study, a novel SOC estimation method based on a combination of temporal convolutional networks (TCN) and gated recurrent units (GRU) is proposed. First, lithium-ion battery data including current, voltage, and temperature are collected under various temperatures and discharge conditions. Then, a hybrid TCN-GRU model is constructed, where TCN is leveraged for its ability to capture long-range temporal dependencies with parallel processing and stable gradients, while GRU is employed for its efficiency in modeling sequential data with fewer parameters and strong temporal memory capabilities. By using current, voltage, and temperature as input features, the model learns the nonlinear relationship between these measurable parameters and SOC, enabling accurate estimation. Finally, the proposed method is validated under multiple temperatures and dynamic load conditions. The experimental results demonstrate that the proposed method achieves SOC estimation errors within 3%, confirming its high accuracy and robustness under varying operating environments.