Anomalies are widely distributed in dam safety monitoring sequences, and their screening and identification is a prerequisite for determining the operational state of a dam. The traditional regression model-based anomaly identification method may cause misjudgment of normal values or omission of anomalies in the monitoring data. To address the above problems, the monitoring data sequence is combined with Kalman filtering method to remove the noise term, and the daily change rate of the measured value is used to replace the denoised data, so as to retain the real evolution trajectory of the data, and then combined with the cloud model, the improved cloud model based on the daily change rate is established. At the same time, a genetic ant colony hybrid algorithm is used to optimise the threshold value of the improved cloud model. The number of outliers before and after denoising and before and after threshold optimisation are compared and analysed respectively. The results show that the overall range of the daily transformation rate is significantly reduced after the raw data has been processed by Kalman filter denoising, while the optimisation of the threshold intervals with the Genetic Ant Colony Hybrid Algorithm shows that its optimised threshold intervals are smaller than the pre-optimisation ones. The results show that the method proposed in this paper can better identify the outliers in the seepage monitoring of dams, reduce the misjudgement caused by noise, and effectively improve the identification accuracy of the outliers.