Skip to contents

Smart charging algorithm

Source: R/smart-charging.R
smart_charging.Rd

This function provides a framework to simulate different smart charging methods (i.e. postpone, interrupt and curtail) to reach multiple goals (e.g. grid congestion, net power minimization, cost minimization). See implementation examples and the formulation of the optimization problems in the documentation website.

Usage

smart_charging(
  sessions,
  opt_data,
  opt_objective,
  method,
  window_days,
  window_start_hour,
  responsive = NULL,
  power_th = 0,
  charging_power_min = 0,
  energy_min = 1,
  energy_max = 1,
  include_log = FALSE,
  show_progress = FALSE,
  lambda = 0
)

Arguments

sessions

tibble, sessions data set containig the following variables: "Session", "Timecycle", "Profile", "ConnectionStartDateTime", "ConnectionHours", "Power" and "Energy".

opt_data

tibble, optimization contextual data. The first column must be named datetime (mandatory) containing the date time sequence where the smart charging algorithm is applied, so only sessions starting within the time sequence of column datetime will be optimized. The other columns can be:

  • static: static power demand (in kW) from other sectors like buildings, offices, etc.

  • import_capacity: maximum imported power from the grid (in kW), for example the contracted power with the energy company.

  • export_capacity: maximum exported power from the grid (in kW), for example the contracted power with the energy company.

  • load_capacity: maximum charging power (in kW) allowed for the flexible EV demand at every time slot. By default the algorithm already limits the optimized setpoint to the nominal power of the connected EVs (the physical charging envelope). This variable can be used to tighten that limit further (e.g. a charging point or transformer power limit).

  • production: local power generation (in kW). This is used when opt_objective = "grid".

  • price_imported: price for imported energy (Euro/kWh). This is used when opt_objective = "cost".

  • price_exported: price for exported energy (Euro/kWh). This is used when opt_objective = "cost".

  • price_turn_down: price for turn-down energy use (Euro/kWh). This is used when opt_objective = "cost".

  • price_turn_up: price for turn-up energy use (Euro/kWh). This is used when opt_objective = "cost".

If columns of opt_data are user profiles names, these are used as setpoints and no optimization is performed for the corresponding user profiles.

opt_objective

character, optimization objective being "none", "grid", "cost", "capacity" or a value between 0 (cost) and 1 (grid). See details section for more information about the different objectives.

method

character, scheduling method being "none", "postpone", "curtail" or "interrupt". If none, the scheduling part is skipped and the sessions returned in the results will be identical to the original parameter.

window_days

integer, number of days to consider as optimization window.

window_start_hour

integer, starting hour of the optimization window.

responsive

Named two-level list with the ratio (between 0 and 1) of sessions responsive to smart charging program. The names of the list must exactly match the Time-cycle and User profiles names. For example: list(Monday = list(Worktime = 1, Shortstay = 0.1))

power_th

numeric, power threshold (between 0 and 1) accepted from setpoint. For example, with power_th = 0.1 and setpoint = 100 for a certain time slot, then sessions' demand can reach a value of 110 without needing to schedule sessions.

charging_power_min

numeric, minimum charging power in kW (default 0). With method = "curtail", a session is never curtailed below it: with charging_power_min = 4, an 11 kW session can be curtailed down to 4 kW and no further. The one exception is a session's last charging slot, where it may charge less than the minimum because it only needs that much to finish (the tail of its charge). A session whose nominal Power is below charging_power_min (e.g. 3.7 kW against 4 kW) is not flexible, whatever the method: it is never responsive, charges as uncontrolled and the setpoint carries its load as fixed.

Up to version 1.12.0 a value below 1 was a share of the nominal power; it is now always kW, so charging_power_min = 0.5 means 0.5 kW.

energy_min

numeric, minimum allowed ratio (between 0 and 1) of required energy. Every session charges at least this share of the energy it requires. The scheduler never curtails, postpones or interrupts a session below it, and the setpoint optimization may drop energy down to it - and no further - when the grid capacity cannot fit energy_max. When even this minimum does not fit, the grid capacity is relaxed only as far as the minimum-energy profile needs, so the minimum is delivered at the expense of the capacity.

energy_max

numeric, maximum allowed ratio (between 0 and 1) of required energy. Every session charges at most this share of the energy it requires, and the setpoint optimization targets exactly this share for the responsive sessions. Lower it to save energy or cost. Must be higher than 0 and at least energy_min.

include_log

logical, whether to output the algorithm messages for every user profile and time-slot

show_progress

logical, whether to output the progress bar in the console

lambda

numeric, penalty on change for the flexible load.

Value

a list with three elements: optimal setpoints (tibble), sessions schedule (tibble) and log messages (list of character vectors, one per window). The date-time values in the log list are in the time zone of the opt_data.

A session whose connection crosses an optimization window boundary is split into one part per window before scheduling (energy shared by connection time), so the sessions schedule carries a Part column and such a session returns several rows per part; its delivered energy is the sum of its rows' Energy grouped by Session.

Details

An important parameter of this function is opt_data, which defines the time sequence of the smart charging algorithm and the optimization variables. The opt_data parameter is directly related with the opt_objective parameter. There are four different optimization objectives implemented by this function:

  • Minimize grid interaction (opt_objective = "grid"): minimizes the peak of the flexible load and the amount of imported power from the grid. If production is not found in opt_data, only a peak shaving objective will be considered.

  • Minimize capacity violations (opt_objective = "capacity"): minimizes only the load slices that exceed the grid capacity limits (import_capacity or export_capacity in opt_data), leaving the rest of the load profile unchanged. Falls back to grid objective when the capacity slice is infeasible.

  • Minimize the energy cost (opt_objective = "cost"): minimizes the energy cost. In this case, the columns grid_capacity, price_imported, price_exported, price_turn_up and price_turn_down of tibble opt_data are important. If these variables are not configured, default values of grid_capacity = Inf, price_imported = 1, price_exported = 0, price_turn_up = 0 and price_turn_down = 0 are considered to just minimize the imported energy.

  • Combined optimization (opt_objective between 0 and 1): minimizes both the net power peaks and energy cost.

  • No optimization (opt_objective = "none"): this will skip optimization and at least one user profile name must be in an opt_data column to be considered as a setpoint for the scheduling algorithm, or a grid capacity variable such as grid_capacity, import_capacityor export_capacity. The user profiles that don't appear in opt_data will not be optimized.

The optimized setpoint of a profile stays within the physical envelope of its responsive sessions: at most their nominal power while they are connected (and load_capacity), and, at every time slot, at least the energy_min share of the energy of the sessions that have already departed. Energy owed to a car leaving early is therefore never postponed into the slots of the cars that stay longer.

Examples

# Example: we will use the example data set of charging sessions
# from the `evsim` package.

# The user profiles of this data set are `Visit` and `Worktime`,
# identified in two different time cycles `Workday` and `Weekend`.
# These two variables in the `sessions` tibble, `Profile` and `Timecycle`,
# are required for the `smart_charging` function and give more versatility
# to the smart charging context. For example, we may want to only coordinate
# `Worktime` sessions instead of all sessions.

# For this example we want the following:
# - Curtail only `Worktime` sessions, which have a responsiveness rate of
# 0.9 (i.e. 90% of Worktime users accept to postpone the session).
# - Minimize the power peak of the sessions (peak shaving)
# - Time series resolution of 15 minutes
# - Optimization window of 24 hours from 6:00AM to 6:00 AM
# - The energy charged can be reduced up to 50% of the original requirement

library(dplyr)

# Use first 50 sessions
sessions <- evsim::california_ev_sessions_profiles %>%
  slice_head(n = 50) %>%
  evsim::adapt_charging_features(time_resolution = 15)
sessions_demand <- evsim::get_demand(sessions, resolution = 15)

# Don't require any other variable than datetime, since we don't
# care about local generation (just peak shaving objective)
opt_data <- tibble(
  datetime = sessions_demand$datetime,
  production = 0
)
sc_results <- smart_charging(
  sessions, opt_data,
  opt_objective = "grid", method = "curtail",
  window_days = 1, window_start_hour = 6,
  responsive = list(Workday = list(Worktime = 0.9)),
  energy_min = 0.5
)