Determination of the optimal number of clusters in harmonic data classification

In many of clustering algorithms, such as K-means and Fuzzy C-mean, the value of the expected numbers of clusters is often needed in advance as an input parameter to the algorithm. Other clustering algorithms estimate this number as the clustering process progresses using various heuristic techniques; however such techniques can also lead to a local minima within the solution space without finding the optimum number of clusters. In this paper, a method has been developed to determine the optimum number of clusters in power quality monitoring data using a data mining algorithm based on the minimum message length technique. The proposed method was tested using data from known number of clusters with randomly generated data points, with data from a simulation of a power system, and with power quality data from an actual harmonic monitoring system in a distribution system in Australia. The results from the tests confirm the effectiveness of the proposed method in finding the optimum number of clusters.