clusterTypicalDaysForOneClass: Generate a set of flow-based typical days on one time period

In rte-antares-rpackage/flowBasedClustering: Cluster 'Flow-Based' Typical Days

Description

Run a clustering algorithm for a given historical period. It creates clusters by gathering the most similar days and chooses among them the best representative: it will be a so-called typical day. The metric used to determine the similarity of two days is a weighted sum of 24 hourly distances, meaning the distances between the domains of the two days at the same hour.

Usage

clusterTypicalDaysForOneClass(dates, vertices, nbCluster, hourWeight = rep(1,
  24), className = NULL, reportPath = getwd(), report = TRUE,
  id_start = 1, maxDomainSize = 10000)

Arguments

dates	character, vector of date (format YYYY-MM-DD).
vertices	data.table, 5 columns : Date : date ( Period : Hour (1:24), period i corresponds to the hourly time-step [i-1 ; i] BE : belgium vertices DE : german vertices FR : french vertices This parameter can be obtained with the function ptdfToVertices.
nbCluster	numeric, number of clusters
hourWeight	numeric, vector of 24 weights, one for each hour of the day. The clustering algorithm will be more accurate for the flow-based domains of the hours with a relatively higher weight.
className	character, name of the class characterizing the studied time period
reportPath	character, path where the report is written. Defaut to getwd()
report	boolean, if TRUE, reports are generated.
id_start	numeric, first identifier of the returned typical days. Default value is 1
maxDomainSize	numeric limit of domain size in each axis. The function will return an error if one domain or more exceed these limits.

Examples

library(data.table)
vertices <- fread(system.file("dataset/vertices_example.txt",package = "flowBasedClustering"))
dates <- seq(as.Date("2016-08-09"), as.Date("2018-09-01"), by = "day")
hourWeight = rep(1, 24)

out1 <- clusterTypicalDaysForOneClass(vertices = vertices,
                                      dates = dates, nbCluster = 2,
                                      className = "myName", id_start = 5, report = FALSE)

out2 <- clusterTypicalDaysForOneClass(vertices = vertices,
                                      dates = dates, nbCluster = 4,
                                      className = "myName2", id_start = 7, report = FALSE)

rbindlist(list(out1, out2))

rte-antares-rpackage/flowBasedClustering documentation built on Nov. 21, 2020, 11:21 a.m.