Toy example IV: Computing reports
Warning
The reporting functionality is still in an early development stage. Beware that major changes might be introduced.
So far, we have worked on scheduling batteries and processes. Now, we are moving to one of the other three pillars of FlexMeasures: reporting.
In essence, reporters apply arbitrary transformations to data coming from some sensors (multiple inputs) and save the results to other sensors (multiple outputs). In practice, this allows to compute KPIs (such as profit and total daily energy production), to apply operations to beliefs (e.g. changing the sign of a power sensor for some time period), among other things.
Note
- Currently, FlexMeasures comes with the following reporters:
PandasReporter: applies arbitrary Pandas methods to sensor data.
AggregatorReporter: combines data from multiple sensors into one using any of the methods supported by the Pandas aggregate function (e.g. sum, average, max, min…).
ProfitOrLossReporter: computes the profit/loss due to an energy flow under a specific tariff.
Moreover, it’s possible to implement your custom reporters in plugins. Instructions for this to come.
Now, coming back to the tutorial, we are going to use the AggregatorReporter and the ProfitOrLossReporter. In the first part, we’ll use the AggregatorReporter to compute the (discharge) headroom of the battery in Toy example II: Adding solar production and limited grid connection. That way, we can verify the maximum power at which the battery can discharge at any point of time. In the second part, we’ll use the ProfitOrLossReporter to compute the costs of operating the process of Tut. Part III in the different policies.
Before getting to the meat of the tutorial, we need to set up up all the entities. Instead of having to do that manually (e.g. using commands such as flexmeasures add sensor), we have prepared a command that does that automatically.
Setup
Just as in previous sections, we need to run the command flexmeasures add toy-account, but this time with a different value for kind:
$ flexmeasures add toy-account --kind reporter
- Under the hood, this command is adding the following entities:
A yearly sensor that stores the capacity of the grid connection.
A power sensor, headroom, to store the remaining capacity for the battery. This is where we’ll store the report.
A ProfitOrLossReporter configured to use the prices that we set up in Tut. Part II.
Three sensors to register the profits/losses from running the three different processes of Tut. Part III.
Let’s check it out!
Run the command below to show the values for the grid connection capacity:
$ TOMORROW=$(date --date="next day" '+%Y-%m-%d')
$ flexmeasures show beliefs --sensor 7 --start ${TOMORROW}T00:00:00+02:00 --duration PT24H --resolution PT1H
Beliefs for Sensor 'grid connection capacity' (ID 7).
Data spans a day and starts at 2023-08-14 00:00:00+02:00.
The time resolution (x-axis) is an hour.
┌────────────────────────────────────────────────────────────┐
│ │
│ │
│ │
│ │
│ │ 1.0MW
│ │
│ │
│ │
│▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀│ 0.5MW
│ │
│ │
│ │
│▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁│ 0.0MW
│ │
│ │
│ │
│ │ -0.5MW
└────────────────────────────────────────────────────────────┘
5 10 15 20
██ grid connection capacity
Moreover, we can check the freshly created source <Source id=6>, which defines the ProfitOrLossReporter with the required configuration. You’ll notice that the config is under the data_generator field. That’s because reporters belong to a bigger category of classes that also contains the Schedulers and Forecasters.
$ flexmeasures show data-sources --show-attributes --id 5
ID Name Type User ID Model Version Attributes
---- ------------ -------- --------- -------------- --------- -----------------------------------------
6 FlexMeasures reporter ProfitOrLossReporter {
"data_generator": {
"config": {
"consumption_price_sensor": 1
}
}
}
Compute headroom
In this case, the discharge headroom is nothing but the difference between the grid connection capacity and the PV power. To compute that quantity, we can use the AggregatorReporter using the weights to make the PV to subtract the grid connection capacity.
In practice, we need to create the config and parameters:
$ echo "
$ {
$ 'weights' : {
$ 'grid connection capacity' : 1.0,
$ 'PV' : -1.0,
$ }
$ }" > headroom-config.json
$ echo "
$ {
$ 'input' : [{'name' : 'grid connection capacity','sensor' : 7},
$ {'name' : 'PV', 'sensor' : 3}],
$ 'output' : [{'sensor' : 8}]
$ }" > headroom-parameters.json
Finally, we can create the reporter with the following command:
$ TOMORROW=$(date --date="next day" '+%Y-%m-%d')
$ flexmeasures add report --reporter AggregatorReporter \
--parameters headroom-parameters.json --config headroom-config.json \
--start-offset DB,1D --end-offset DB,2D \
--resolution PT15M
Now we can visualize the headroom in the following link, which should resemble the following image.
The graph shows that the capacity of the grid is at full disposal for the battery when there’s no sun (thus no PV generation), while at noon the battery can only discharge at 280kW max.
Process scheduler profit
For the second part of this tutorial, we are going to use the ProfitOrLossReporter to compute the losses (defined as cost - revenue) of operating the process from Tut. Part III, under the three different policies: INFLEXIBLE, BREAKABLE and SHIFTABLE.
In addition, we’ll explore another way to invoke reporters: data generators. Without going too much into detail, data generators create new data. The thee main types are: Reporters, Schedulers and Forecasters. This will come handy as the three reports that we are going to create share the same config. The config defines the price sensor to use and sets the reporter to work in losses mode, which means that it will return costs as positive values and revenue as negative values.
Still, we need to define the parameters. The three reports share the same structure for the parameters with the following fields:
input: sensor that stores the power/energy flow. The number of sensors is limited to 1.
output: sensor to store the report. We can provide sensors with different resolutions to store the same results at different time scales.
Note
It’s possible to define the config and parameters in JSON or YAML formats.
After setting up config and parameters, we can invoke the reporter using the command flexmeasures add report.
The command takes the data source id, the files containing the parameters and the timing parameters (start and end).
For this particular case, we make use of the offsets to indicate that we want the report to encompass the day of tomorrow.
Inflexible process
Define parameters in a JSON file:
$ echo "
$ {
$ 'input' : [{'sensor' : 4}],
$ 'output' : [{'sensor' : 9}]
$ }" > inflexible-parameters.json
Create report:
$ flexmeasures add report --source 6 \
--parameters inflexible-parameters.json \
--start-offset DB,1D --end-offset DB,2D
Check the results here. The image should be similar to the one below.
Breakable process
Define parameters in a JSON file:
$ echo "
$ {
$ 'input' : [{'sensor' : 5}],
$ 'output' : [{'sensor' : 10}]
$ }" > breakable-parameters.json
Create report:
$ flexmeasures add report --source 6 \
--parameters breakable-parameters.json \
--start-offset DB,1D --end-offset DB,2D
Check the results here. The image should be similar to the one below.
Shiftable process
Define parameters in a JSON file:
$ echo "
$ {
$ 'input' : [{'sensor' : 6}],
$ 'output' : [{'sensor' : 11}]
$ }" > shiftable-parameters.json
Create report:
$ flexmeasures add report --source 6 \
--parameters shiftable-parameters.json \
--start-offset DB,1D --end-offset DB,2D
Check the results here. The image should be similar to the one below.
Now, we can compare the results of the reports to the ones we computed manually in this table). Keep in mind that the report is showing the profit of each 15min period and adding them all shows that it matches with our previous results.