Power grids are undergoing significant changes due to the increase in renewable energy sources and the large- scale introduction of electric vehicles, impacting Power Quality significantly. Consequently, network operators conduct extensive measurement campaigns, generating large amounts of data that contain valuable information about disturbance characteristics. This paper presents a methodology for typical daily patterns detection in PQ parameters using an unsupervised k-means clustering method. Approaches for determining the suitable number of clusters and handling trends, seasonality and weekly cycles are proposed Recognizing the typical patterns allows for the study of similarities and differences between measurement sites, monitoring PQ conditions at specific sites, and detecting changes in emission behavior over time. This information enhances PQ management, enabling network operators to respond promptly to changes in disturbance patterns. The proposed methodology has been tested on long-term field measurements recorded at different measurement sites of 110 kV network supplying large cities in China, demonstrating its effectiveness and adaptability to dynamic nature of PQ measurement data.