Background Perinatal depression is common worldwide, which can cause many adverse effects on the physical and mental health of the mother and baby, as well as the whole family. The Edinburgh Postnatal Depression Scale (EPDS) is an efficient and effective instrument for perinatal depression. However, few studies have examined its longitudinal measurement invariance (LMI) during the whole perinatal period, which is particularly important in longitudinal studies, such as exploring developmental trajectories of perinatal depression and evaluating the effects of certain interventions. Methods 4139 pregnant women from 24 hospitals in 15 provinces of China were measured using EPDS in the first, second, third trimesters and 6 weeks postpartum. Exploratory factor analysis and confirmatory factor analysis were used to explore the factor structure of EPDS at each time point. Multi-group analyses were performed to examine LMI of EPDS. Results A three-factor model was optimal at all time points, showing the clearest factor structure and best model fit: Anhedonia (Items 1-2), Anxiety (Items 3-6), Depression (Items 7-10). Internal reliability of EPDS was good at all time points (e.g., Cronbach's alpha > 0.80). A series of multi-group analyses further indicated that the EPDS held strict LMI (configural, metric, scalar and strict invariance) during the perinatal period. Conclusion The findings further confirmed three-factor structure and good reliability of the EPDS when used in Chinese pregnant and postpartum women. The LMI justified comparisons of EPDS scores among different measurement time points.