Parametric simulation using eplurs

[divyanshusood]

Hi,

I want to ask a few questions regarding 'eplusr' in R. I want to perform parametric simulations with parameters such as:

Orientation [0:45:360]
Solar Heat Gain Coefficient [0:0.1:1]
Heating Setpoint [18:1:21] and so on.

How can I perform this using 'eplusr' in R?

Once this is done and I have the dataset, I plan to perform sensitivity analysis to determine the influential parameters impacting various outputs such as:

Zone-wise annual heating consumption
Zone-wise annual electricity consumption
Zone-wise annual CO2 concentration

Is there a way I can create a single csv file with all ranges of input parameters and corresponding output using eplur?

Please guide me with some recommendations. Thanks for your help.

[hongyuanjia]

Hi @divyanshusood. eplusr has a dedicated vignette about parametric simulation. I would like you to first read through it to get a whole picture of how to perform parametric simulations. Basically you wrap your steps of modifying the IDF in a function (I call it a measure in eplusr) and pass it to ParametricJob$apply_measure().

Is there a way I can create a single csv file with all ranges of input parameters and corresponding output using eplur?

Yes, you can. Let's say func is your measure function which takes three arguments, i.e. orientation (orient), SHGC (shgc) and Heating setpoint (hgsp). You can do whatever you want to generate values for those arguments, of cause, including reading from an external CSV file. Just like something below:

func <- function(orient, shgc, hgsp) {
    ...
}

csv <- read.csv(PATH_CSV)

param$apply_measure(
    func,
    orient = csv$orient,
    shgc = csv$shgc,
    hgsp = csv$hgsp
)

[divyanshusood]

Hi Hongyuanjia,
I hope you're doing fine.
I am using the following code to add occupancy from csv to IDF file but getting an error every time.

> create_schedules_for_all_profiles(idf_base_path, file_path, new_names_househousehold)
Error in `map()`:
ℹ In index: 41.
Caused by error in `fread()`:
! skip=5 but the input only has 2 lines
Run `rlang::last_trace()` to see where the error occurred.

_code below this line
The aim of this code is to add multiple schedules in to IDF file. This code will take first set of profiles from each csv files and then add it int IDF file and save it. and again do the process to take second set of profile and add into IDF file and safe new IDF. the csv files are in a folder, the name of csv files are accom_apt_month1_dayFriday, for each month there are 7 files each for one weekday. so in total 84 files, 12 months * 7 weekdays =84 files.

library(eplusr)
library(purrr)
library(data.table)
library(lubridate)
library(hms)

read_sch_data <- function(path, profile_col, na = 0.0) {
  meta <- fread(path, nrows = 5)
  month <- as.integer(meta[1, 2])
  num <- as.integer(meta[2, 2])
  type <- meta[3, 2]
  weekday <- meta[4, 2]
  
  # Read the actual schedule data, skipping metadata
  data <- fread(path, skip = 5)
  # Get column names dynamically
  col_names <- names(data)
  
  # Ensure the correct column is selected based on profile_col index
  if (length(col_names) < profile_col) {
    stop("Profile column index exceeds number of columns in data.")
  }
  
  time_col_name <- col_names[1]  # First column for time
  profile_col_name <- col_names[profile_col]  # Dynamically choose profile column
  
  # Extract and rename selected columns
  selected_data <- data[, .(Time = get(time_col_name), Profile = get(profile_col_name))]
  selected_data[, Time := as.character(Time)]
  selected_data[, Time := as_hms(Time)]
  
  # Calculate interval assuming times are in 'HH:MM:SS' format
  interval <- as.numeric(diff(as.POSIXct(selected_data$Time[1:2], format = "%H:%M:%S"))) / 60
  
  return(list(month = month, num = num, type = type, weekday = weekday, interval = interval, data = selected_data))
}


create_sch_weekly <- function(idf, name, list_days) {
  WEEKDAY <- c("Sunday", "Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday" )
  # stop if any weekday data is missing
  if (any(mis_wd <- !WEEKDAY %in% names(list_days))) {
    stop("Missing schedule for weekday [", paste(WEEKDAY[mis_wd], collapse = ", "), "]")
  }
  
  # create a "AlwaysOff" for weekends, design days and other days
  # adjust this as you need
  if (!idf$is_valid_class("Schedule:Day:Interval") ||
      !idf$is_valid_name("AlwaysOff", "Schedule:Day:Interval")) {
    idf$add(Schedule_Day_Interval = list("AlwaysOff", "On/Off", "No", "Until: 24:00", 0))
    
  }
  # create a daily schedule
  idf$add(
    Schedule_Week_Daily = list(
      name,
      list_days$Sunday$name(),
      list_days$Monday$name(),
      list_days$Tuesday$name(),
      list_days$Wednesday$name(),
      list_days$Thursday$name(),
      list_days$Friday$name(),
      list_days$Saturday$name(),
      "AlwaysOff",
      "AlwaysOff",
      "AlwaysOff",
      "AlwaysOff",
      "AlwaysOff"
    )
  )[[1L]]
}

create_sch_yearly <- function(idf, name, list_months,  type_limits = "Any number") {
  # get the start and end day for each month
  start <- lubridate::ymd("2022-01-01") + months(0:11)
  end <- lubridate::rollforward(start)
  
  # combine start day, end day and month schedules
  data <- data.table::data.table(
    index = seq(12),
    start = start, end = end,
    sch = purrr::map(list_months, ~.$name())
  )
  
  # get the start and end day field value
  data[, `:=`(
    start_month = as.list(lubridate::month(start)),
    start_day = as.list(lubridate::mday(start)),
    end_month = as.list(lubridate::month(end)),
    end_day = as.list(lubridate::mday(end))
  )]
  
  # remove original start and end column since not needed
  data[, `:=`(start = NULL, end = NULL)]
  
  # melt the data based on index column
  m <- data.table::melt.data.table(data, "index")
  data.table::setorderv(m, "index")
  
  idf$add(
    Schedule_Year = as.list(c(name, type_limits, m$value))
  )[[1L]]
}

# Function to create weekly and yearly schedules in the IDF
create_sch_weekly <- function(idf, name, list_days) { ... }
create_sch_yearly <- function(idf, name, list_months, type_limits = "Any number") { ... }

# Function to create schedules for all profiles
create_schedules_for_all_profiles <- function(idf_base_path, file_path, new_names) {
  sample_file <- list.files(file_path, pattern = "\\.csv$", full.names = TRUE)[1]
  sample_data <- fread(sample_file, nrows = 1)
  num_profiles <- ncol(sample_data) - 1
  
  for (profile_col in 2:(num_profiles + 1)) {
    idf <- read_idf(idf_base_path)
    sch_data <- map(list.files(file_path, pattern = "accom_apt_month[0-9]+_day.*\\.csv$", full.names = TRUE), read_sch_data, profile_col = profile_col, na = 0)
    
    # Save a new IDF for each profile set
    idf_save_path <- sprintf("%s/Dublin_Apartment_Profile_Set_%d.idf", file_path, profile_col - 1)
    idf$save(path = idf_save_path, copy_external = TRUE)
  }
}


idf_base_path <- "~/Documents/RQ-3_journal/DeterministicAnalysis/EnergyPlus_Models/Apartment/Apartment_all_idf/Dublin_Apartment_4occ.idf"
file_path <- "~/Documents/RQ-3_journal/DeterministicAnalysis/newtrial/1occ/apt"
new_names_household <- c("jan_house", "feb_house", "mar_house", "april_house", "may_house", "june_house", "july_house", "aug_house", "sept_house", "oct_house", "nov_house", "dec_house")

create_schedules_for_all_profiles(idf_base_path, file_path, new_names_househousehold)

[hongyuanjia]

  # Make a deep copy of the default IDF file for each set
  idf <- copy(idf_default)

You cannot copy an IDF like that. If you want to copy an IDF, you should do idf_default$clone().

[divyanshusood]

It worked, thank you so much.

[hongyuanjia]

Good to know it worked.

[divyanshusood]

Hi Hongyuanjia,

I hope you're going well. And thank you for all your previous support.

I need some guidance/suggestion from you. I am developing a Bayesian Neural Network (BNN) based probabilistic occupancy model so that I can determine the uncertainties associated with occupancy in estimating energy consumption at various scales. I have the BNN model (in Python) and now I want to integrate that with EnergyPlus so that I can draw a sample from Posterior distribution and use those samples to perform the EnergyPlus simulations using IDF file.

I need your suggestion on how can I link the BNN model with EnergyPlus so that I can plot the distribution of energy consumption. Any suggestion would be appreciated.

Thank you
Divyanshu

[hongyuanjia]

That's too genral question. If your model output an annual value, you may use it as a Schedule:File object.

Btw, this thread is mainly used to discuss your questions about parametric simulations using eplusr. If you want to discuss other topic, please open a new thread instead of continuning commenting on this one.