---
title: "EV model object"
output: rmarkdown::html_document
vignette: >
  %\VignetteIndexEntry{EV model object}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(evprof)
```

The EV model object defined by `{evprof}` is generated with function `get_ev_model()`. This function returns an object of class `evmodel`. This object prints a summary of its components. The package provides an example of `evmodel` created in the [California study case article](https://resourcefully-dev.github.io/evprof/articles/california.html), using the charging sessions data provided by [ACN](https://acnportal.readthedocs.io/en/latest/).

```{r ev_model}
evprof::california_ev_model
```

The `evmodel`object has two components:

* `metadata`: creation date, data time zone, if the scale of connection/energy models is natural or logarithmic, ...
* `models`: tibble containing the different time-cycles models and information. The columns of this tibble are:

    * `time_cycle`: character, given name to the time-cycle
    * `months`: integer vector, corresponding months of the time-cycle
    * `wdays`: integer vector, corresponding days of the time-cycle (week starting on day 1)
    * `user_profiles`: tibble with every user profile GMM models. The columns of this tibble are:
    
        * `profile`: character vector, given name to the user profile
        * `ratio`: numeric, share of daily sessions corresponding to this profile
        * `connection_models`: tibble with the connection bi-variate GMM
        * `energy_models`: tibble with the energy uni-variate GMM
        

The model itself is composed by multiple Gaussian models (GMM). The `connection_models` are Gaussian models of two variables (connection start time and connection duration) and the `energy_models` are built with a single variable (energy). Thus, the statistic features of `connection_models` are a centroid ($\mu$), a covariance matrix ($\Sigma$) and the weight or ratio of the corresponding model. Besides, the statistic features of `energy_models` are a mean ($\mu$), a standard deviation ($\sigma$) and the weight or ratio of the corresponding model.


Let's take a look to these statistical features of the Worktime user profile for Working days in the California model:
```{r}
california_ev_model$models
```
```{r}
workday_model <- california_ev_model$models$user_profiles[[1]]
workday_model
```

```{r}
worktime_model <- workday_model[2, ]
```

The connection model is a mixture of 3 bi-variate Gaussian Models:
```{r}
worktime_model$connection_models
```


On the other hand, the energy models can be based on the charging rate (`Power` variable) of the sessions. It has been observed that the energy demand increases together with the charging power (big vehicles have larger batteries and can charge at higher rates). Thus, function `get_energy_models` has the logical parameter `by_power` to fit the Energy Gaussian Models for the different groups of charging powers separately. In the case of California study case, we set `by_power=FALSE`, that's why we got the `Unknown` in the `energy_models` tibble with a `ratio` of 1:

```{r}
worktime_model$energy_models[[1]]
```

Thus, the energy model of all Worktime sessions is a mixture of 9 Gaussian models:

```{r}
worktime_model$energy_models[[1]]$energy_models[[1]]
```