The invention belongs to the field of civil engineering structure health monitoring, and provides a structure monitoring data anomaly identification method based on multivariate statistic analysis. The method includes the steps: firstly, building a multivariate statistic analysis model for normal monitoring data of a structure; secondly, converting an abnormal identification process of the monitoring data into a statistical hypothesis testing problem in a noise subspace of the model; thirdly, solving the statistical hypothesis testing problem to derivate a new statistic for identifying abnormal monitoring data; finally, determining a reasonable threshold value of the statistic, and judging a fact that anomalies exist in the monitoring data when the statistic exceeds the threshold value.