smartOASYS

Overview

This package implements methodologies for smart meters data.

Installation

#install the package
devtools::install_github("aefdz/smartOASYS")

#load the package
library(smartOASYS)

Test usage

Load the example data and plot one day for the 22 inverters.

data("solarPlant1")


matplot(solarPlant1$raw0[,1,], type = "l", main = "Raw")

matplot(solarPlant1$smooth0[,1,], type = "l", main = "Zero derivative smooth")

matplot(solarPlant1$smooth1[,1,], type = "l", main = "First derivative smooth")

High dimensional outlier detection

Compute time series of depths and scaled depths

data <- aperm(solarPlant1$smooth0, c(3,1,2))

depthsTime <- timeDepth(data, typeDepth = "FMD")

Compute the prototype, the distances to the prototype and the outlier detection rule

# Depths
timeSeriesDepth_mat <- matrix(unlist(depthsTime[1,]), ncol = dim(data)[1], nrow = dim(data)[3])
colnames(timeSeriesDepth_mat) <- dimnames(data)[[1]]

# Scaled Depths
#timeDepth_mat <- matrix(unlist(depthsTime[2,]), ncol = dim(data)[1], nrow = dim(data)[3])
#colnames(timeDepth_mat) <- dimnames(data)[[1]]

Plot the time series of depths

matplot(timeSeriesDepth_mat, type = "l", xlab = "Day", main = "22 Panels")

# Prototype. Low alpha to illustrate some outlier
outliers <- evolutionOutlier(timeSeriesDepth_mat, alpha = 0.5)

matplot(timeSeriesDepth_mat, type = "l", xlab = "Day", main = "22 Panels", col = "black")
matplot(timeSeriesDepth_mat[,outliers$out_trimmed_mean_MC], type = "l", add = TRUE, col = "red", lwd = 2)

References

Elías, A., Morales. J. M. and Pineda, S. (2022). A high dimensional functional time series approach to evolution outlier detection for grouped smart meters. Quality Engineering.