Electrical consumption and renewable profile clusterization based on k-medoids method

The planning of renewable energy systems has become a widely studied topic in the scientific literature; for this, the authors use annual historical data to determine if a system is feasible from various points of view that can be technical, economic, or environmental. The large amount of data that is used can make studies computationally expensive and time-consuming. This work develops a novel methodology that allows to overcome these problems presented in classical methodologies. To achieve this goal, this chapter presents data processing and uncertainty management techniques, as measured data may contain inaccuracies and outliers, which are generally caused by untimely incidents, unexpected events, or device failures. Subsequently, to reduce the large amount of data, a clustering technique was used through a temporal representation based on a set of selected representative days; for this, the k-medoids method was used to obtain the representative days of the available measurements. In this way, the total number of representative days that must be considered to obtain accurate results is much less than the total number of scenarios required by other techniques.