This paper presents enhanced Self-Organizing Maps (SOMs) for exploratory multivariate time series analysis in the context of temporal data mining. The main idea lies in an adequate combination of approaches with SOMs for temporal processing. It is part of a recently developed method that introduces several abstraction levels for temporal knowledge conversion. The method provides a conversion of discovered temporal patterns in multivariate time series with enhanced SOMs into a linguistic knowledge representation, in form of temporal grammatical rules. This method was successfully applied to a problem in medicine. Even some previously unknown knowledge was found.