The collection of nuclear power plants operating data is the basis for subsequent fault diagnosis and obtaining the operating status of nuclear power plants, but equipment failures and external interference will lead to missing operating monitoring data, which will reduce the quality of the data and thus reduce the accuracy of the subsequent analysis results. To solve this problem, this paper utilizes the fact that nuclear power plants have accumulated a large amount of operational data and researches the method of generating adversarial imputation network (GAIN)–based imputation method for missing values of nuclear power plants’ operational data. The generator in the model estimates the missing values by learning the distribution of the true values, and the discriminator in the model discriminates which values are true and which are generated with the help of a hint matrix. The hint reveals partial information about the missing original samples to the discriminator, which the discriminator uses to focus its attention on the quality of the imputation of particular data values. Finally, a training set and a test set were constructed for comparative experiments on the PCTran simulation platform by simulating the operational data of the AP1000 as an example. The experimental results demonstrate that the investigated algorithm achieves lower root mean square error (RMSE), verifying the feasibility and accuracy of the method.