R/tpk-piv_iv_week.R

In SunsVoc: Constructing Suns-Voc from Outdoor Time-Series I-V Curves

# Handle function to obtain pseudo I-V curve and other predicted IV features
# at reference conditions given a week (or more) of data

#' Obtain Psuedo IV Curve and other predicted IV features at reference conditions for
#' a given week, used internally in IVxbyx function.
#'
#' @param df A dataframe containing timeseries I-V features of one period.
#' @param temp The reference module temperature to correct the Psuedo IV curve to be,
#' unit of the temperature should be Celsius.
#' @param N_c Number of cells in series. Equal to the total number of cells in the system.
#' @param isc_1sun (optional) Input an Isc 1-Sun value manually. Leave NULL to have one generated
#' from the dataframe.
#'
#' @return Psuedo-IV Curve data with features extracted and evaluation parameters
#' of fitting grouped for a single, given period.
#'
#' @examples
#' df <- read_df_raw(df_wbw,0.02,7)
#' df_slice <- dplyr::filter(df, df$n_period == 1)
#' # Check that this has enough data! needs more than 10 rows to be meaningful
#' nrow(df_slice)
#' # needs median temperature
#' temp <- median_temp(df_wbw)
#' res <- p_iv.week(df_slice, temp = temp, N_c = 60)
#'
#'
#' @importFrom rlang .data
#' @importFrom magrittr %<>%
#' @import dplyr
#' @export
p_iv.week <- function(df, temp, N_c, isc_1sun = NULL){

  # create new cols for fitting
  df <- df %>%
    dplyr::filter(.data$isc > 1e-3)

  # The physical model
  df <- df %>% mutate(T_K = .data$modt + 273.15,
                      lnSun = 1.38e-23/1.6e-19 * N_c * .data$T_K * log(.data$isc),
                      isc2 = .data$isc^2,
                      T_lnIsc2 = .data$T_K * .data$lnSun * log(.data$isc),
                      rs = as.numeric(.data$rs),
                      expVoc = exp(-.data$voc/.data$T_K) / .data$isc,
                      I0 = (N_c * 1.38e-23 * .data$T_K) / (1.6e-19 * .data$isc)) # scaling for lm fit

  # select data that has non nan or inf Tln(Ir) values
  df %<>% filter(!(is.na(.data$lnSun) | is.infinite(.data$lnSun)))

  if (is.null(isc_1sun)) {
    # obtain 1sun isc
    isc_1sun <- isc.1sun(df$isc, df$poa)
  }

  # correct voc by temperature and construct pseudo I-V curve
  voc_mod <- voc.corr(df)
  # replace all temperature to desired temperature as new predictor
  df_STC <- df %>%
    mutate(modt = temp)

  # correct voc to given at various irradiance level according to weekly model
    voc_corr <- predict(voc_mod, newdata = df_STC)
    mod_res <- summary(voc_mod)
    rmse <- sqrt(mean(voc_mod$residuals ^ 2))

    # obtain predicted voc value at 1000 W/m2 (with 1 sun isc)
    # and given temperature
    df_piv <- data.frame(voc_corr = voc_corr,
                         isc = df$isc)

  # modeling Imp and Vmp
    mod_imp <- stats::lm(imp ~ isc:modt + isc2:modt + isc + isc2, data = df)
    mod_vmp <- stats::lm(vmp ~ lnSun + T_lnIsc2 + modt, data = df)
    # modeling Rs
    mod_rs <- stats::lm(rs ~ I0, data = filter(df, .data$isc > 2 & .data$isc < 10))

    ln_1Sun = 1.38e-23/1.6e-19*N_c*(temp+273.15)*log(isc_1sun)
    #predict imp and vmp for 1sun and given temp
    imp_fit <- stats::predict(mod_imp,
                       newdata = data.frame(isc = isc_1sun, isc2 = isc_1sun^2, modt = temp))
    vmp_fit <- stats::predict(mod_vmp,
                       newdata = data.frame(lnSun = ln_1Sun,
                                            modt = temp,
                                            T_lnIsc2 = (temp + 273.15) * ln_1Sun * log(isc_1sun)
                                            ))
    res_vmp <- summary(mod_vmp)
    vmp_rmse <- sqrt(mean(mod_vmp$residuals ^ 2))
    pmp_fit <- imp_fit * vmp_fit
    res_rs <- summary(mod_rs)
    rs_rmse <- sqrt(mean(mod_rs$residuals ^ 2))
    names(imp_fit) <- "imp_fit"
    names(vmp_fit) <- "vmp_fit"
    names(pmp_fit) <- "pmp_fit"

  # Setting names attribute
    voc_1sun <- stats::predict(voc_mod,
                        newdata = data.frame(lnSun = ln_1Sun,
                                             modt = temp))
    names(voc_1sun) <- "voc_1sun"
    rs_fit <- stats::predict(mod_rs,
                      newdata = data.frame(I0 = N_c * 1.38e-23 * (temp + 273.15)/(1.6e-19 * isc_1sun)))
    names(rs_fit) <- "rs_fit"

  # Calculate ideality factor, create res
    n <- voc_mod$coefficients[2]
    names(n) <- "n"
    names(isc_1sun) <- "isc_1sun"

    res <- c(df_piv, isc_1sun, voc_1sun, n,
             imp_fit, vmp_fit, pmp_fit, rs_fit,
             "voc_adjR2" = mod_res$adj.r.squared,
             "voc_rmse" = rmse,
             "vmp_adjR2" = res_vmp$adj.r.squared,
             "vmp_rmse" = vmp_rmse,
             "rs_adjR2" = res_rs$adj.r.squared,
             "rs_rmse" = rs_rmse)

  return(res)
}