3.1. Execution Model¶
The execution of CO2MPAS model for a single vehicle is a stepwise procedure
of 3 stages:
These are invoked repeatedly, and subsequently combined, for the various cycles,
as shown in the “active” flow-diagram of the execution, below:
The models in the diagram are nested; explore by clicking on them.
- Precondition: identifies the initial state of the vehicle by running
a preconditioning WLTP cycle, before running the WLTP-H and WLTP-L
The inputs are defined by the
input.precondition.wltp_pnode, while the outputs are stored in
- Calibration: the scope of the stage is to identify, calibrate and select
(see next sections) the best physical models from the WLTP-H and WLTP-L
input.calibration.wltp_x). If some of the inputs needed to calibrate the physical models are not provided (e.g.
initial_state_of_charge), the model will select the missing ones from precondition-stage’s outputs (
output.precondition.wltp_p). Note that all data provided in
input.calibration.wltp_xoverwrite those in
- Prediction: executed for the NEDC and as well as for the WLTP-H and
WLTP-L cycles. All predictions use the
calibrated_models. The inputs to predict the cycles are defined by the user in
input.prediction.xxxnodes. If some or all inputs for the prediction of WLTP-H and WLTP-L cycles are not provided, the model will select from
`output.calibration.wltp_xnodes a minimum set required to predict CO2 emissions.
3.2. Excel input: data naming conventions¶
This section describes the data naming convention used in the CO2MPAS template
.xlsx file). In it, the names used as sheet-names, parameter-names
and column-names are “sensitive”, in the sense that they construct a
data-values tree which is then fed into into the simulation model as input.
These names are split in “parts”, as explained below with examples:
base.input.precondition.WLTP-H.ts └┬─┘ └─┬─┘ └────┬─────┘ └─┬──┘ └┬┘ scope────────┘ │ │ │ │ usage──────────────┘ │ │ │ stage───────────────────────┘ │ │ cycle─────────────────────────────────┘ │ sheet_type──────────────────────────────────┘
First 4 parts above are optional, but at least one of them must be present on a sheet-name; those parts are then used as defaults for all parameter-names contained in that sheet. type is optional and specify the type of sheet.
plan.target.prediction.initial_state_of_charge.WLTP-H └┬─┘ └─┬─┘ └────┬────┘ └──────────┬──────────┘ └──┬─┘ scope(optional)─┘ │ │ │ │ usage(optional)───────┘ │ │ │ stage(optional)────────────────┘ │ │ parameter────────────────────────────────────────┘ │ cycle(optional)──────────────────────────────────────────────────┘
OR with the last 2 parts reversed:
plan.target.prediction.WLTP-H.initial_state_of_charge └──┬─┘ └──────────┬──────────┘ cycle(optional)─────────────────────────┘ │ parameter──────────────────────────────────────────────┘
- The dot(
.) may be replaced by space.
- The usage and stage parts may end with an
s, denoting plural, and are not case-insensitive, e.g.
3.2.1. Description of the name-parts¶
base[default]: values provided by the user as input to CO2MPAS.
plan: values selected (see previous section) to calibrate the models and to predict the CO2 emission.
flag: values provided by the user as input to
meta: values provided by the user as meta data of the vehicle test.
input[default]: values provided by the user as input to CO2MPAS.
data: values selected (see previous section) to calibrate the models and to predict the CO2 emission.
output: CO2MPAS precondition, calibration, and prediction results.
target: reference-values (NOT USED IN CALIBRATION OR PREDICTION) to be compared with the CO2MPAS results. This comparison is performed in the report sub-model by
config: values provided by the user that modify the
wltp-pis specified as cycle]: data related to the precondition stage.
calibration[default]: data related to the calibration stage.
nedcis specified as cycle]: data related to the prediction stage.
selector: data related to the model selection stage.
nedc-h: data related to the NEDC High cycle.
nedc-l: data related to the NEDC Low cycle.
wltp-h: data related to the WLTP High cycle.
wltp-l: data related to the WLTP Low cycle.
wltp-precon: data related to the preconditioning WLTP cycle.
wltp-p: is a shortcut of
nedc[default]: is a shortcut to set values for both
wltp[default]: is a shortcut to set values for both
all: is a shortcut to set values for
param: any data node name (e.g.
vehicle_mass) used in the physical model.
sheet_type: there are three sheet types, which are parsed according to their contained data:
- pl [parsed range is
#A1:__]: table of scalar and time-depended values used into the simulation plan as variation from the base model.
- pa [parsed range is
#B2:C_]: scalar or not time-depended values (e.g.
- ts [parsed range is
#A2:__]: time-depended values (e.g.
gears). Columns without values are skipped. COLUMNS MUST HAVE THE SAME LENGTH!
- ..note:: If it is not defined, the default value follows these rules:
When scope is
plan, the sheet is parsed as pl. If scope is
baseand cycle is missing in the sheet-name, the sheet is parsed as pa, otherwise it is parsed as ts.
- pl [parsed range is
3.3. Calibrated Physical Models¶
There are potentially eight models calibrated from input scalar-values and time-series (see API Reference):
- engine_speed_model, and
Each model is calibrated separately over WLTP_H and WLTP_L. A model can contain one or several functions predicting different quantities. For example, the electric_model contains the following functions/data:
These functions/data are calibrated/estimated based on the provided input (in the particular case: alternator current, battery current, and initial SOC) over both cycles, assuming that data for both WLTP_H and WLTP_L are provided.
co2_params model has a third possible calibration configuration
(so called ALL) using data from both WLTP_H and WLTP_L combined
(when both are present).
3.4. Model selection¶
Since v1.4.1-Rally, this part of the model remains disabled,
flag.use_selector is true.
For the type approval mode the selection is fixed. The criteria is to select the models calibrated from WLTP_H to predict WLTP_H and NEDC_H; and from WLTP_L to predict WLTP_L and NEDC_L.
While for the engineering mode the automatic selection can be enabled adding -D flag.use_selector=True to the batch command. Then to select which is the best calibration (from WLTP_H or WLTP_L or ALL) to be used in the prediction phase, the results of each stage are compared against the provided input data (used in the calibration). The calibrated models are THEN used to recalculate (predict) the inputs of the WLTP_H and WLTP_L cycles. A score (weighted average of all computed metrics) is attributed to each calibration of each model as a result of this comparison.
The overall score attributed to a specific calibration of a model is the average score achieved when compared against each one of the input cycles (WLTP_H and WLTP_L).
For example, the score of electric_model calibrated based on WLTP_H when predicting WLTP_H is 20, and when predicting WLTP_L is 14. In this case the overall score of the the electric_model calibrated based on WLTP_H is 17. Assuming that the calibration of the same model over WLTP_L was 18 and 12 respectively, this would give an overall score of 15.
In this case the second calibration (WLTP_L) would be chosen for predicting the NEDC.
In addition to the above, a success flag is defined according to upper or lower limits of scores which have been defined empirically by the JRC. If a model fails these limits, priority is then given to a model that succeeds, even if it has achieved a worse score.
The following table describes the scores, targets, and metrics for each model: