The invention discloses a method and a system for regulating the unbalance of a sample set type of transient voltage stability of a power network, belonging to the field of power system stability analysis and evaluation. The method of the invention takes the history sample set collected from the power network history record as the main component of the training sample set, generates the predictionsample by using the time domain simulation of the anticipated fault set under the dispatching operation plan, and adopts the Euclidean distance comparison among different types of samples to test thequalification of the prediction sample. Combining the predicted samples and the historical samples, the training sample set for mining and learning is constructed, which can alleviate and adjust theimbalance of the historical sample set. Decision tree algorithm is used to classify and learn the training sample set, and the decision tree model is obtained. The decision tree model is used as the transient voltage stability assessment model in the real-time monitoring process.