With a world needing to progress to a more renewable-based energy mix, wave renewable is indispensable to achieve this goal. Studies showed that it is necessary to cartography wave energy converter and location pairs as varying wave climates require different devices. However, there is not one but many designs of the converters. Hence, pre-matching converter classes with wave climates would enable such computational-demanding global cartography. Then, devices would be matched to these classes to find where they suit most. Power production potentials display features to be naturally pre-classified: (1) presence of a plateau of same maximal value, and (2) associated capture width and possible linear reduction to a 1-dimensional curve; and different power-representations are considered. Then, the principal component analyses are employed to determine the new power production potential classification. This method enables to cluster into a simple 3-dimensional space distributions of data characterised by many dimensions. After calculating the Euclidian distance between the projected data to determine the clusters, different technics are assessed to obtain the representative classes. Results highlighted the non-correlation between previous classifications and the power production potential. The 46 considered power production potentials formed 16 classes. Finally, guidelines are provided to use this new classification.