Custom evmodel

Simplifying our model

The evmodel object obtained with evprof contains all statistic parameters of the Gaussian Mixture Models for both connection and energy models (see this article). However, these GMM can be quite complex with a lot of clusters (mixtures) and the JSON file describing the model can result in a large file.

To check the complexity of the models and describe all parameters within the JSON file, the function get_evmodel_parameters provides the model parameters in a list form:

ev_model <- evsim::california_ev_model
get_evmodel_parameters(ev_model)

## $Workday
## $Workday$Visit
## $Workday$Visit$ratio
## [1] 0.4603
## 
## $Workday$Visit$connection_models
## # A tibble: 3 × 5
##   start_mean start_sd duration_mean duration_sd ratio
##        <dbl>    <dbl>         <dbl>       <dbl> <dbl>
## 1      12.2      1.01          4.73        1.09 0.235
## 2       7.44     1.04          4.66        1.23 0.446
## 3      15.3      1.01          2.45        1.25 0.319
## 
## $Workday$Visit$energy_models
## # A tibble: 1 × 4
##   charging_rate energy_mean energy_sd ratio
##   <chr>               <dbl>     <dbl> <int>
## 1 Unknown              11.8      1.18     1
## 
## 
## $Workday$Worktime
## $Workday$Worktime$ratio
## [1] 0.5397
## 
## $Workday$Worktime$connection_models
## # A tibble: 3 × 5
##   start_mean start_sd duration_mean duration_sd ratio
##        <dbl>    <dbl>         <dbl>       <dbl> <dbl>
## 1       7.31     1.00          9.75        1.01 0.305
## 2       7.31     1.03          8.90        1.03 0.428
## 3       6.55     1.01          9.64        1.00 0.267
## 
## $Workday$Worktime$energy_models
## # A tibble: 1 × 4
##   charging_rate energy_mean energy_sd ratio
##   <chr>               <dbl>     <dbl> <int>
## 1 Unknown              16.1      1.14     1
## 
## 
## 
## $Weekend
## $Weekend$Visit
## $Weekend$Visit$ratio
## [1] 1
## 
## $Weekend$Visit$connection_models
## # A tibble: 2 × 5
##   start_mean start_sd duration_mean duration_sd ratio
##        <dbl>    <dbl>         <dbl>       <dbl> <dbl>
## 1       7.54     1.01          7.23        1.07 0.205
## 2      11.3      1.06          3.55        1.30 0.795
## 
## $Weekend$Visit$energy_models
## # A tibble: 1 × 4
##   charging_rate energy_mean energy_sd ratio
##   <chr>               <dbl>     <dbl> <int>
## 1 Unknown              13.3      1.49     1

Even though the California EV model doesn’t have a lot of clusters the composition of the model is complex enough, and the logarithmic scale makes it difficult to interpret. Therefore, the evsim package also provides a function to simplify the clusters calculating the average connection and energy pattern for every user profile, converting the logarithmic values to the natural scale:

get_evmodel_summary(ev_model)

## $Workday
## # A tibble: 2 × 8
##   profile  ratio start_mean start_sd duration_mean duration_sd energy_mean
##   <chr>    <dbl>      <dbl>    <dbl>         <dbl>       <dbl>       <dbl>
## 1 Visit    0.460      11.1      1.02          3.97        1.20        11.8
## 2 Worktime 0.540       7.11     1.02          9.36        1.02        16.1
## # ℹ 1 more variable: energy_sd <dbl>
## 
## $Weekend
## # A tibble: 1 × 8
##   profile ratio start_mean start_sd duration_mean duration_sd energy_mean
##   <chr>   <int>      <dbl>    <dbl>         <dbl>       <dbl>       <dbl>
## 1 Visit       1       10.5     1.05          4.30        1.25        13.3
## # ℹ 1 more variable: energy_sd <dbl>

Now we can see that every time-cycle is defined by a simple user profile with an average behaviour. This format is also useful to save the model in Excel, since it can be saved directly in an Excel file where every time-cycle will be a different worksheet:

get_evmodel_summary(ev_model) %>% 
  writexl::write_xlsx("ev_model.xlsx")

Designing a custom model

Moreover, sometimes maybe it is desired to build a custom model with specific average connection start time and duration and average energy charged. In this case, the Excel file created with the previous command can be used as a template to define our custom user profiles and parameters. After editing the file we can read it again:

custom_model <- purrr::map(
  readxl::excel_sheets("ev_model_custom.xlsx") %>% purrr::set_names(),
  ~ readxl::read_excel("ev_model_custom.xlsx", sheet = .x)
)

Then, these parameters can be passed to argument parameters_lst of function evsim::get_custom_ev_model, where every element of this list must be a different time-cycle (as our object custom_model is). In the evsim::get_custom_ev_model function also the months and weekdays corresponding to every time-cycle (worksheet in the Excel file) and the time-zone of the use case must be configured in parameters months_lst, wdays_lst and data_tz, respectively. Finally, connection_log and energy_log parameters may be always FALSE since it is assumed that a custom model is built in the natural scale:

evmodel_custom <- evsim::get_custom_ev_model(
  names = names(custom_model),
  months_lst = list(1:12, 1:12),
  wdays_lst = list(1:5, 6:7),
  parameters_lst = custom_model,
  connection_log = F,
  energy_log = F,
  data_tz = "America/Los_Angeles"
)