plr_bootstrap_uncertainty {PVplr}

Bootstrap: Resampling data going into each Model

Description

This function determines the uncertainty of a PLR measurement through resampling data for each model, prior to putting the data through the model.

Usage

plr_bootstrap_uncertainty(
  df,
  n,
  fraction = 0.65,
  var_list,
  model,
  by = "month",
  power_var = "power_var",
  time_var = "time_var",
  data_cutoff = 100,
  np = NA,
  pred = NULL
)

Arguments

df

A dataframe containing pv data.

n

(numeric) Number of samples to take. The higher the n value, the longer it takes to complete, but the results become more accurate as well.

fraction

The fraction of data that constitutes a resample for the bootstrap.

var_list

A list of variables obtained through plr_variable_check.

model

the String name of the model to bootstrap. Select from:

• "xbx" (plr_xbx_model),

• "correction" (plr_xbx_utc_model),

• "pvusa" (plr_pvusa_model),

• or "6k" (plr_6k_model).

by

String, either "day", "week", or "month". Time over which to perform plr_yoy_regression.

power_var

Variable name of power in the dataframe. This must be the variable's name after being put through your selected model. Typically power_var

time_var

Variable name of time in the dataframe. This must be the variable's name after being put through your selected model. Typically time_var

data_cutoff

The number of data points needed to keep a value in the final table. Regressions over less than this number and their data will be discarded.

np

The system's reported name plate power. See plr_6k_model.

pred

passed to predict_data in model call. See plr_xbx_model for an example.

Value

Returns PLR value and uncertainty calculated with bootstrap of data going into power correction models

Examples

# build var_list


var_list <- plr_build_var_list(time_var = "timestamp",
                               power_var = "power",
                               irrad_var = "g_poa",
                               temp_var = "mod_temp",
                               wind_var = NA)
# Clean Data
test_dfc <- plr_cleaning(test_df, var_list, irrad_thresh = 100,
                         low_power_thresh = 0.01, high_power_cutoff = NA)
                         
xbx_mbm_plr_uncertainty <- plr_bootstrap_uncertainty(test_dfc, n = 2, 
                                                     fraction = 0.65, by = 'month',
                                                     power_var = 'power_var', time_var = 'time_var',
                                                     var_list = var_list, model = "xbx",
                                                     data_cutoff = 10, np = NA,
                                                     pred = NULL)