Load Profile Segmentation for Electricity Market Settlement

Data Analytics for Profiling Low-Voltage Customers with Smart Meter Readings

“…The kmeans algorithm was used to segment households according to the electricity pattern while the household characteristics were studied to determine the influences on the electricity consumptions. In [24], load profile segmentation was developed for market settlement purposes. The hierarchical clustering approaches with Euclidean and dynamic time waring distances and average and Ward's linkage criteria were compared offline for analyzing measured consumption profiles of residential, commercial, and industrial consumers.…”

“…The dataset size used by the surveyed papers spans from the yearly profile of one industrial consumer only [14] to a few tens of two customer categories (residential and small enterprises users) [22], to hundreds of consumers without category characterization [15] or residential only [24,25], to a few thousand (i.e., 2771 in [20], 5000 in [17]), to a few tens of thousands of residential customers only [16,23]. Other studies refer to load profiles of aggregated customers, i.e., at level of primary substations (few hundreds in [19]), or at level of secondary substations (tens of thousands in [21]), or referred to buildings (tens in [27], thousands in [26]).…”

“…Moreover, considering the load profiling approaches in the reviewed literature, even if more than one clustering technique is exploited in some papers, the use of one clustering technique only for the load profiling process is common [14][15][16][17][18][20][21][22][23][24][25]. In [19] and [27], two different clustering techniques were used in a sequential way, while in [26], three different algorithms were compared offline.…”

“…Two reviewed papers aimed at enhancing the performances [15] or providing a comparative analysis [14] of acknowledged algorithms. The papers [16,17,22,23] addressed the topic of identifying common patterns among one category of customers only, while more than one category was studied in [20,[24][25][26][27]. In this paper the electric behavior of all the categories of Italian LV customers is studied (i.e., industrial, commercial, agricultural, and residential).…”

“…Once a typical pattern is identified for the analyzed profiles in the dataset, the load profiling activity poses the problem of assigning these typical load profiles to customers that do not belong to the analyzed dataset. This stage is not addressed in [14][15][16][17]19,21,22,24,27]. In [18], Markov chains were proposed for creating synthetic models of the MV/LV and HV/MV interfaces to be used for analyzing the network at the higher voltage level.…”