The performance and service life of ultracapacitors are highly dependent on accurate modeling and state-of-charge (SOC) estimating. To overcome the model parameter errors caused by the various temperatures and different SOC intervals, the H infinity filter (HIF) is employed to estimate the ultracapacitor SOC based on a variable temperature model. For the application of the HIF method, the Thevenin model is first developed with a small terminal voltage estimation error. Then, the model parameters are optimally identified using the ant colony optimization (ACO) algorithm. Next, a variable temperature model is established to improve the adaptability of the ultracapacitor model, and the HIF is utilized for the ultracapacitor SOC estimation. Finally, to verify the performance of the variable temperature model and the proposed SOC estimation method, a series of experiments are conducted. The analysis results illustrate that the mean absolute error (MAE) of the SOC estimation values based on the variable temperature model is decreased by 39.62% compared to the one based on the nonvariable temperature model. Meanwhile, the proposed state estimation scheme based on the variable temperature model is accurate with estimation values maximum error (ME) and root-mean-squared error (RMSE) less than 0.80% and 0.60%, respectively. The HIF-based SOC estimation method also shows a good robustness with a short convergence time within 90.00 s when the SOC initial error is set to 0.20.