PV power clustering as a means to evaluate energy storage options

The rapid advancement of Renewable Energy Sources and especially Photovoltaics (PV) was aided by the generous Feed-in-Tariffs employed worldwide. However, such incentives are disappearing and the need of market revitalization is apparent. Providing energy storage in PV installations is considered as an attractive option from a technical point of view, since it may help reduce uncertainties and make such power source dispatchable. However, energy storage is not attractive to investors yet. In fact, several factors should be taken into consideration when performing a techno-economic analysis of an energy storage option in a PV installation. Such factors include, for example, the minimum allowable State Of Charge (SOC) and capital costs related to energy and power capacity. In this paper, we propose to incorporate clustering techniques when storage options are evaluated in a PV installation. With this approach, the basic input to calculations (i.e. the PV power output) is not simulated from archive meteorological data, but the actual power produced is taken into account for existing PVs. Historical data of power production from a PV station are grouped into clusters with representative power curves called centroids. In this way, evaluating storage options becomes less time consuming and simpler. This paper takes real data from a PV installation and formulates clusters using several techniques. After a comparison of the methods used, a discussion of how such clustering formulation may be used for optimization problems is presented.