Failures in ice accretion on power transmission lines during winter and early spring will lead to serious power system issues. This paper innovatively incorporates weather and topographic factors often overlooked due to their non-quantifiable nature. Textual weather and topographic data are quantified and combined with meteorological factors, including temperature, humidity, wind speed, and wind direction, to construct feature vectors. A transmission line ice accretion prediction model is developed using the SVM algorithm and 5-fold cross-validation, employing predicted ice accretion probability as the model output. According to the comparison results of the model before and after feature optimization, the model’s performance improves after incorporating weather and terrain conditions. The precision, recall, accuracy, and AUC reach 94.85%, 80.63%, 88.05%, and 0.9729, respectively, higher than the former’s 85.71%, 78.75%, 82.70%, and 0.9341. Field investigations are also implemented to validate the algorithm, and the results achieved by the proposed algorithm significantly align with ground observation data. Therefore, it is of great significance to prevent the occurrence of power grid icing accidents by incorporating weather and terrain factors into the ice accretion prediction model.