Time series data is a crucial component of numerous real-life applications, with various models designed for tasks such as regression, classification, forecasting, and missing value imputation. However, the clustering of time series based on their intrinsic features remains underexplored. In this thesis, we propose a method for time series clustering. Our approach involves unsupervised learning of representations for time series data and subsequently refining these representations through an iterative process between clustering assignments and updates of deep neural networks. We evaluate the proposed method using both academic datasets and practical datasets, specifically the BMW Snapshot time series. The results demonstrate the effectiveness and applicability of our approach in real-world scenarios.