In this article, a case study is presented on applying cluster analysis techniques to evaluate the level of power quality (PQ) parameters of a virtual power plant. The conducted research concerns the application of the K-means algorithm in comparison with the agglomerative algorithm for PQ data, which have different sizes of features. The object of the study deals with the standardized datasets containing classical PQ parameters from two sub-studies. Moreover, the optimal number of clusters for both algorithms is discussed using the elbow method and a dendrogram. The experimental results show that the dendrogram method requires a long processing time but gives a consistent result of the optimal number of clusters when there are additional parameters. In comparison, the elbow method is easy to compute but gives inconsistent results. According to the Calinski–Harabasz index and silhouette coefficient, the K-means algorithm performs better than the agglomerative algorithm in clustering the data points when there are no additional features of PQ data. Finally, based on the standard EN 50160, the result of the cluster analysis from both algorithms shows that all PQ parameters for each cluster in the two study objects are still below the limit level and work under normal operating conditions.