Configuration of a Simulator

The Simulator of a model works as our interface to communicate between the orchestrator of the simulation and the model entities. This means that the simulator for the PV system model sends input values to the PV systems, initiates the model calculations, and collects output values to return to the orchestrator (e.g., for sending the power values to the connected energy management systems). As there may be multiple entities of a model connected to a single simulator, you only need one simulator of a model type for a simulation, as this simulator communicates with all entities of its model type. Hence, a _simulator.py is always paired with a _model.py.

This page explains the structure of a _simulator.py helping you to create a simulator for your own model.

Note

All code-blocks derive from charging_station_simulator.py as of (01/24) if not stated otherwise.

Import mosaik and model

Import of the mosaik simulator base class mosaik_api_v3 and the corresponding model + additional imports

import mosaik_api_v3
from eelib.core.devices.charging_station import charging_station_model
import eelib.utils.validation as vld
from copy import deepcopy

---

Listing of model META

State of simulator (whether it is time-discrete or event-based/hybrid), see the glossary

META = {
   "type": "hybrid",

Listing of ALL attributes for each modeltype: At first, all attributes of the model class…

   "models": {
      "ChargingStation": {
         "public": True,
         "params": ["init_vals"],
         "attrs": [
            "p",
            "step_size",
            "time",
            ...

… input attributes also listed in "trigger" list …

         "trigger": [
            "p_set",
            "ev_data",
         ],  # input attributes

… output attributes also listed in ""non-persistent" list.

         "non-persistent": [
            "p",
            "p_min",
            "p_max",
            "cs_data",
         ],  # output attributes

---

Initialization of simulator class

Constructor __init__ for this simulator and the initialization function init()

class Sim(mosaik_api_v3.Simulator):
    """Simulator class for eELib charging station model."""

    def __init__(self):
        """Constructs an object of the Charging-Station:Sim class."""

        super(Sim, self).__init__(META)

        ...

    def init(self, sid, scenario_config, time_resolution=1.0):
        """Initializes parameters for an object of the Charging-Station:Sim class."""

        ...

---

Following are the so-called core functions create(), step() and get_data() that are found in every _simulator.py.

Caution

All core functions of the simulators are called by mosaik and need to be existent!

---

Creation of model entities in create()

Creation of model entities by assigning ID and name, creating it by its model type, and storing the entity data

def create(self, num, model_type, init_vals):
    """Creates entities of the eELib charging station model."""

    # generated next unused ID for entity
    next_id = len(self.entities)

    # create empty list for created entities
    entities_orchestrator = []

    for i in range(next_id, next_id + num):
        # create entity by specified name and ID
        eid = "%s%s%d" % (model_type, "_", i)
        full_id = self.sid + "." + eid

        # get class of specific model and create entity with init values after validation
        entity_cls = getattr(charging_station_model, model_type)
        vld.validate_init_parameters(entity_cls, init_vals[i])
        entity = entity_cls(
            eid,
            **init_vals[i],
            step_size=self.scenario_config["step_size"],

        # add info to the simulators entity-list and current entities
        self.entities[eid] = {
            "eid": eid,
            "etype": model_type,
            "model": entity,
            "full_id": full_id,
        }
        entities_orchestrator.append({"eid": eid, "type": model_type})

    return entities_orchestrator

---

Stepping of models in step()

Within the step() method, inputs from mosaik are set into the model entity attributes they are intended for. Afterwards, the entities are called and their calculation is executed with the models’ step() method.

This example is coming from the storage_simulator.py (01/24).

Take input data and set values of entities, then call the model calculation

# assign property values for each entity and attribute with entity ID
# process input signals: for the entities (eid), attr is a dict for attributes to be set
for eid, attrs in inputs.items():
    # for the attributes, setter is a dict for entities with corresponding set values
    for attr, setter in attrs.items():
        # set values corresponding to its type
        ...

# call step function for each entity in the list
for eid, entity_dict in self.entities.items():
    entity_dict["model"].step(time)

Note

You might return the next timestep for when this model should be called again. This is needed for time-based models (e.g. when reading data from a csv file, the next value should be read and send out) and optional for event-based models (e.g. the adaption of energy levels for a storage model is needed, but a heatpump could maybe just be triggered by a new set value).

---

Handling of output data in get-data

From a defined set of output properties given by the outputs parameter from mosaik, the values of the transmitting entities are read and stored into a data dict that is returned

for transmitter_ename, attrs in outputs.items():
    # get name for current entity and create dict field
    entry = self.entities[transmitter_ename]
    if transmitter_ename not in self.output_cache:
        self.output_cache[transmitter_ename] = {}

    # loop over all targeted attributes and check if info is available
    for attr in attrs:
        if attr not in self.meta["models"][type(entry["model"]).__name__]["attrs"]:
            raise ValueError("Unknown output attribute: %s" % attr)

        # create empty field for cache and output data
        if attr not in self.output_cache[transmitter_ename]:
            self.output_cache[transmitter_ename][attr] = {}

        output_data_to_save = getattr(entry["model"], attr)

        # check if this value changed
        ...

    # check if nothing is to be send out - send output 1 step later to avoid waiting for data
    if not flag_output_changed:
        if self.timestep == self.scenario_config["n_steps"] - 1:  # is last timestep?
            data["time"] = self.timestep + 1
        else:
            data = {"time": self.timestep}

    return data

For each timestep, the output data is continuously stored and compared to the lastly sent data (output_cache) so if nothing new is to be sent out, only the timestep will be sent. In case of the last timestep, we need to adjust the signal by transmitting not the current but the next timestep to avoid a loop of endless calling between the simulators.