“…In a broad sense, we have managed to reproduce both the results in [12], where large dispersions of hosting capacities inside clusters were observed when using numbers of clusters in the range of 10 to 30, and the results of [8], where relatively low levels of constraint type purity were observed when using numbers of clusters below 10. However, it should be noted that the quality measurements used in those studies (boxplot-based Fig.…”
Section: E Results Interpretationmentioning
confidence: 94%
“…However, it should be noted that the quality measurements used in those studies (boxplot-based Fig. 5: Quality measurement of constraint type prediction in [12] and cluster purity in [8]) only aim at measuring the level of homogeneity inside clusters, while the measurements introduced here aim at measuring the quality of the prediction provided by the representative feeders. As a consequence, these can be used to select a number of clusters (and thus representative networks) to use in a particular study as a function of the level of accuracy expected for the study.…”
Section: E Results Interpretationmentioning
confidence: 99%
“…By building on the works in [8] and [12], we propose two external quality measures that can be computed for each number of clusters :…”
In this paper, we will assess the relevance of applying hierarchical agglomerative clustering algorithms on medium voltage feeder descriptive parameters (average line impedance, capacity, total length, etc.) in order to select representative feeders for long term planning studies. To achieve this, we start by creating a dataset of medium voltage feeders (bus and line geometries and characteristics) by combining domain knowledge with open datasets of distribution network layouts, district level electricity consumptions and building footprints. We then use this dataset to calculate descriptive attributes for each feeder as well as their maximal load hosting capacity and the associated type of constraint. Afterwards we perform a statistical analysis of the descriptive attributes in order to select the most relevant to use as inputs of the clustering algorithms. Finally, we apply a hierarchical agglomerative clustering algorithm with varying number of clusters, assess the quality of the results using internal and external validation and evaluate the ability of the medoids of each cluster to represent the behavior of the corresponding feeders.
“…In a broad sense, we have managed to reproduce both the results in [12], where large dispersions of hosting capacities inside clusters were observed when using numbers of clusters in the range of 10 to 30, and the results of [8], where relatively low levels of constraint type purity were observed when using numbers of clusters below 10. However, it should be noted that the quality measurements used in those studies (boxplot-based Fig.…”
Section: E Results Interpretationmentioning
confidence: 94%
“…However, it should be noted that the quality measurements used in those studies (boxplot-based Fig. 5: Quality measurement of constraint type prediction in [12] and cluster purity in [8]) only aim at measuring the level of homogeneity inside clusters, while the measurements introduced here aim at measuring the quality of the prediction provided by the representative feeders. As a consequence, these can be used to select a number of clusters (and thus representative networks) to use in a particular study as a function of the level of accuracy expected for the study.…”
Section: E Results Interpretationmentioning
confidence: 99%
“…By building on the works in [8] and [12], we propose two external quality measures that can be computed for each number of clusters :…”
In this paper, we will assess the relevance of applying hierarchical agglomerative clustering algorithms on medium voltage feeder descriptive parameters (average line impedance, capacity, total length, etc.) in order to select representative feeders for long term planning studies. To achieve this, we start by creating a dataset of medium voltage feeders (bus and line geometries and characteristics) by combining domain knowledge with open datasets of distribution network layouts, district level electricity consumptions and building footprints. We then use this dataset to calculate descriptive attributes for each feeder as well as their maximal load hosting capacity and the associated type of constraint. Afterwards we perform a statistical analysis of the descriptive attributes in order to select the most relevant to use as inputs of the clustering algorithms. Finally, we apply a hierarchical agglomerative clustering algorithm with varying number of clusters, assess the quality of the results using internal and external validation and evaluate the ability of the medoids of each cluster to represent the behavior of the corresponding feeders.
“…More recent work [11] has investigated the accuracy of clustering distribution system for determining hosting capacity. While it has been shown that clustering does not necessarily perfectly group all feeders into narrow hosting capacity bands, clustering does provide a nice method to select a subset of feeders that represent the range of all feeders.…”
the Utility Application Review and Approval process for interconnecting distributed energy resources to the distribution system. Currently this process is the most time-consuming of any step on the path to generating power on the distribution system [1]. This CSI RD&D solicitation three project has completed the tasks of collecting data from the three utilities, clustering feeder characteristic data, detailed modeling of 16 representative feeders, and analysis of PV impacts to those feeders. In this report, gaps and limitations in the current screening process-California Rule 21 [2] are identified. Technically-based suggestions to improve Rule 21 are made along with a validation of those methods. Industry Challenge Various incentive programs have increased the number of solar PV system interconnection requests to levels never before seen. Utilities must evaluate these interconnection requests to ensure proper operation of the grid is maintained. To assist utilities in quickly evaluating these systems, certain "screens" have been developed over the years that help identify when issues may or may not arise. The most common screening method takes into account the ratio of solar PV to peak load (15%), however it does not take into account the locational impact of PV nor the feeder-specific characteristics that can strongly factor in to whether issues may occur. EPRI has shown that a feeder's hosting capacity for accommodating PV is strongly determined by location of PV as well as a specific feeder's characteristics [3].
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.