Abstract: In temporal databases, the amount of data increases rapidly because of the introduction of time dimension which in turn causes storage abnormity and data redundancy. Therefore, it’s very important to effectively design temporal database. Data dependency is a core concept in database design, analogously temporal data dependency is also a core concept in temporal database design. Besides the storage redundancy and update abnormity associated with temporal functional dependencies, there also exist the storage redundancy and update abnormity associated with temporal multi-valued dependencies in temporal databases. However, because of temporal multi-valued dependencies’ complexity, normal forms decomposition based on it can’t always be lossless. To resolve the problem of lossless decomposition of temporal schemes with multi-valued dependencies, the following work has been done. The relativity between temporal multi-valued dependencies and lossless decomposition in strong total temporal schemes is investigated. The concepts of regular temporal multi-valued dependency and time granularity of attribute set are given. A set of effective inferring rules of regular temporal multi-valued dependencies is presented and theories of relativity between temporal multi-valued dependencies and lossless decomposition in strong total temporal schemes are also presented. The problem of lossless decomposition of temporal schemes with multiple time granularities under conditions of regular temporal multi-valued dependencies is resolved, laying the foundation for further normalization of temporal schemes with regular temporal multi-valued dependencies.