Accurate estimation of the State of Charge (SOC) of new energy vehicle batteries is crucial for accurately predicting the required power supply for emergency power supply situations and achieving optimal scheduling of emergency power sources and loads in microgrids. This article proposes a wide temperature open circuit voltage model based on support vector machine regression (SVR). This model can accurately reflect the performance changes of batteries under different temperature conditions, thereby significantly improving the stability, robustness, and initial error correction ability of SOC estimation. By combining the second-order RC equivalent circuit model, dynamic forgetting factor recursive least squares (DFFRLS), and adaptive square root unscented Kalman filter (ASRUKF), accurate estimation of SOC has been achieved. In the case of an initial error of 20%, this method demonstrated excellent adaptability and accuracy in testing over a wide temperature range (including extreme temperatures), with root mean square error (RMSE) and mean absolute error (MAE) both below 1.95%, while the complexity of the algorithm did not significantly increase.