Pipeline

What Does the SunsVoc package do?

With the SunsVoc package, Isc-Voc curves can be constructed with outdoor time-series I-V curves of photovoltaic (PV) modules instead of having to be measured in the lab. Suns-Voc (or Isc-Voc) curves can provide the current-voltage (I-V) characteristics of the diode of photovoltaic cells without the effect of series resistance. Time series of four different power loss modes; namely, uniform current, recombination, series resistance, and current mismatch, can be calculated based on obtained Isc-Voc curves. The details for the analysis method can be found in Wang et al. 2020.

Loading the package

library(SunsVoc)

The Data Format

The imported data should have the following columns: a string containing the IV curve for each timestamp, extracted features from the ddiv package’s IVFeatures function run on individual IV curves, Voc and Isc readings from any source (if available), Plane of Array (POA) irradiance and module temperature (in Celsius) readings, and a timestamp. If you have Global Horizontal Irradiance (GHI) instead of POA, convert it first to POA. This can be done using the Python package PVLib, using the latitude, longitude, and elevation parameters of the installation site.

Use the following variable names in order to ensure the package processes your data correctly:

Deprecated Name Column name Unit/Standard Format (if applicable)
Time stamp tmst Local yyyy-mm-dd hh:mm:ss
Module temperature modt \(^\circ\)C –
Plane of array irradiance poa W/m2 –
Open circuit voltage voc V –
Short circuit current isc A –
Voltage at max. power vmp V –
Current at max. power imp A –
Max. power point pmp W –
Series resistance rs \(\Omega\) –

These are the key variable names. Look at the example data below to see other variable names. The extracted features obtained with ddiv should have the correct names when they are extracted.

Note that the timestamp in each column must match the IV curve with it.

# below is the function used to read in our csv
# df_wbw <- read_df_raw_from_csv('../../data/sa42589-albsf_full_ddiv.csv', 0, 7)

colnames(df_wbw)
#>  [1] "tmst" "ivdf" "modt" "poa"  "isc"  "voc"  "rs"   "pmp"  "imp"  "vmp"

Extracting I-V Features with the ddiv Package

Use the ddiv package’s IVfeatures function to extract features from individual IV curves. See package documentation for ddiv on how to use this function (note that ddiv requires the segmented package, among others). Note that ddiv returns a list, so it’s a good idea to pass its results to a dataframe.

# Load required packages
library(dplyr)
library(purrr)
library(magrittr)
library(ddiv)

# I-V curve feature extraction using the IVfeature function
str_ddiv <- function(iv_str, pb) {
  iv <- char_to_df(iv_str)
  # initialize result for when IVfeature encounters an error
  err_df <- data.frame("Isc" = NA, "Rsh" = NA,
                       "Voc" = NA, "Rs" = NA,
                       "Pmp" = NA,"Imp" = NA,
                       "Vmp" = NA, "FF" = NA)

  # calculate ddiv with error trapping
  res <- tryCatch(
    as.data.frame(IVfeature(I = iv$I, V = iv$V, num = 200, crt = 0.1, crtvalb = 0.2)),
    error = function(e) err_df
  )
  pb$tick()$print()
  return(res)
}

# Batch processing of I-V curves for feature extraction
batch_ddiv <- function(df) {
  iv_list <- df$ivdf 
  pb <- dplyr::progress_estimated(length(iv_list))
  ddiv_df <- iv_list %>% map_dfr(~str_ddiv(iv_str = ., pb))
  res <- cbind(df$tmst, ddiv_df)
  return(res)
}

# It is recommended to test on a small number of I-V curves 
# to tweak the input parameters of the IVfeature function
test <- sample_n(df_wbw, 5)
ddiv_test <- batch_ddiv(test)
colnames(ddiv_test) <- c("tmst", "isc", "rsh", "voc", "rs", "pmp", "imp", "vmp", "ff" )

# Bind the ddiv result with the original dataframe
test_res <- full_join(test[,c(1,3:4)], ddiv_test[,c(1:2,4:8)])

# See the column names of the extracted features
colnames(test_res)

Prerequisites for the Power Loss Mode Analysis

It is helpful to know the median temperature of the studied PV module prior to the analysis. The df_wbw is the example dataset in the package. A method for median temperature determination is included as follows:

T_corr <- median_temp(df_wbw)

T_corr
#> [1] 30

It is recommended that the analysis period be chosen so that each psuedo-IV curve has at least 300 points. In order to achieve this, make sure that there are at least 300 Isc and Voc pairs in each period. 10-minute intervals of collection is generally good for 7 days.

Lastly, the power_loss_bat function, which deals with power loss modes, requires a subset of the larger dataframe. Use the select_init_df function to do this. The input dataframe for both select_init_df and IVXbX functions would be better to be filtered according to the current accuracy of the tracing equipment and have the rows with NAs in the modt and pmp variables removed.

# Read the raw data with tracer accuracy for filtering and time period for pseudo I-V curves
df <- read_df_raw(df_wbw, tracer_accuracy = 0.02, t_period = 7)

# Subset the data for the first 21 days
df_init <- select_init_df(df, days = 21) 

# See the column names of the initial dataframe
colnames(df_init)
#>  [1] "tmst"     "ivdf"     "modt"     "poa"      "isc"      "voc"     
#>  [7] "rs"       "pmp"      "imp"      "vmp"      "date"     "day"     
#> [13] "n_period"

Psuedo-IV Curve Generation

Psuedo-IV Curve generation is handled by a master function called IVXbyX which calls a variety of subfunctions.

# Psuedo-IV curve generation 
df_full <- IVXbyX(df, corr_temp = T_corr, N_c= 60)

# See the column names of the generated data
colnames(df_full)
#>  [1] "date"      "day"       "n_period"  "piv"       "pisc"      "pvoc"     
#>  [7] "prsh"      "prss"      "ppmp"      "pimp"      "pvmp"      "pfff"     
#> [13] "isc_1sun"  "voc_1sun"  "n"         "imp_fit"   "vmp_fit"   "pmp_fit"  
#> [19] "rs_fit"    "voc_adjR2" "voc_rmse"  "vmp_adjR2" "vmp_rmse"  "rs_adjR2" 
#> [25] "rs_rmse"

Power Loss Modes

The power_loss_phys_bat function is finally used to calculate the power loss modes: losses due to uniform current, recombination, series resistance, and current mismatch.

# Power loss mode calculation
res <- power_loss_phys_bat(df_full, df_init, corr_T = T_corr, N_c = 60) 

# Generate an example table for power loss modes
knitr::kable(res[1:5, ], caption = "Power Loss Modes")
Power Loss Modes
date uni_I rec rs I_mis
2019-03-01 -19.880795 -3.1387682 -17.87311 22.425457
2019-03-08 6.228080 0.0547520 -12.79985 -5.088667
2019-03-15 9.483216 -0.1758531 -16.13820 -1.318198
2019-03-22 6.686206 1.0209962 -13.69075 -4.007085
2019-03-29 7.101225 0.4053336 -14.52323 19.873580

Visualization of the Power Loss Modes

Visualization of the result can be done using the ggplot2 package.

library(ggplot2)
ggplot(data = res, aes(x = date)) +
  geom_point(aes(y = uni_I, color = "Uniform current", shape = "Uniform current"), size = 2) +
  geom_point(aes(y = rec, color = "Recombination",  shape = "Recombination"), size = 2) +
  geom_point(aes(y = rs, color = "Rs loss", shape = "Rs loss"),  size = 2) +
  geom_point(aes(y = I_mis, color = "I mismatch", shape = "I mismatch"), size = 2) +
  ylab(expression(paste(Delta, "Power (W)"))) + xlab("Date") +
  theme_bw() +
  theme(axis.text = element_text(size = 12), axis.title = element_text(size = 12),
        legend.text = element_text(size = 10), legend.position = "top") +
  scale_shape_manual(values = c(4, 8, 17, 16), name = "Power loss \n mode") +
  scale_colour_discrete(name = "Power loss \n mode")