Time Series Clustering and Evaluation of Unknown Working Conditions of Mismatched Photovoltaic Array Systems

It is not clear whether unknown working conditions of a solar power system yield distinct clusters of currents and voltages where each cluster has definite physical implication. The problem is associated with smartness of next generation of condition monitoring, prognosis and fault diagnosis in the field of photovoltaic power systems. Raw time data of currents and voltages are divided into many subsequences to measuring similarity through DTW, then cluster similarity through the k-means method. All clusters are evaluated by a purity value approach. It is shown that all derived clusters agree well with certain group or groups of unknown working conditions.