本文在分析我国国债期货收益率特征的基础上构建了基于结构突变的16种HAR族模型并用于对我国国债期货已实现波动率进行预测,同时对预测模型进行样本内和样本外预测效果比较。研究结果表明,国债期货收益率存在明显结构变化;投机活动变量可以为国债期货已实现波动率预测提供重要信息;无论在样本内还是样本外,所构建的基于结构突变的预测模型均较标准HAR-RV有更好的预测效果,而且所构建的HAR-CJ类模型具有最好的样本外预测能力。 Based on the analysis of the characteristics of bond yields in our country, this paper constructs 16 HAR family models based on structural mutation and uses them to predict the realized volatility of China’s national debt futures. At the same time, the comparison between the in-sample and the out-of-sample prediction results is made. The results show that there are obvious structural changes in the yield of the national debt futures. Speculative variables can provide important information for the realized volatility forecast of the national debt futures. The model based on the structural abrupt change is better than the standard HAR -RV has better prediction results, and the constructed HAR-CJ class model has the best ability of out-of-sample prediction.