Stochastic structure tutorial

Welcome to Spine Toolbox's Stochastic System tutorial.

This tutorial provides a step-by-step guide to get started with the stochastic structure. More information can be found in the documentation on the stochastic structure. It is recommended to make sure you are able to get the simple system tutorial working first.

In this tutorial we will take a look at independent scenarios and stochastic paths.

Setup starting from simple system tutorial

We create a new Spine Toolbox project and start from the simple system tutorial.

For the Spine Toolbox project

• Open Spine Toolbox
• Create a new Spine Toolbox project
• Add two data store items (input and output)
  • set the dialect to sqlite
  • push the new database button
• Add the run SpineOpt tool
  • connect the databases to the SpineOpt tool
  • in the properties pane of the SpineOpt tool,
  move the available resources to the tool arguments

For the simple system tutorial

• Download the simple system database (json file)

from the examples folder in the SpineOpt repository (you can save the json file in your Spine Toolbox project folder)

• Enter the input database such that you are in the spine db editor
• Go to the hamburger menu (Alt+F) and select import
• Locate the downloaded file to import the simple system
• We save our results when we commit to the database,

so go again to the hamburger menu and select commit. The update message can be something like this: import simple system tutorial.

Independent scenarios

Recall from the simple system tutorial that there actually already is a stochastic structure present. Let us take a closer look at that structure.

The scenarios are the labels that are available to the user to label their data. Don't worry, we'll come back to that later. Here, there is currently one scenario realization.

The scenarios are managed by the stochastic structure. Foremost, the stochastic structure is connected to the model with the model__stochastic_structure relationship. The stochastic structure is also connected to different parts of the energy system to manage the stochastic structure in these parts. With the model__defaultstochasticstructure relationship we can connect the scenario to the entire energy system. Here, there is one stochastic structure deterministic which is also the systems default.

It is quite simple to add an independent scenario to this existing stochastic structure.

• Add a scenario object and call it 'independent'
• Add a stochastic_structure__stochastic_scenario relationship between independent and deterministic

either from the tree view (right click -> new relationship) or from the graph view (right click -> add relationship)

Now we can use these labels in the values for the energy system.

• Change the demand parameter at the electricity_node from 150.0 to a map

(right click -> edit, parameter type map)

• for the x column we can use our scenario labels, for the Value column we can choose our values
• Choose realization 150.0 and independent 100.0
• Save/Commit the results

That is it! We can now run the model and the output database will show the results for both scenarios. In the realization scenario power plant b produces an output of 50. In the independent scenario power plant b does not produce anything as the demand is low enough for power plant a to produce all the necessary energy.

Stochastic path

SpineOpt always works with stochastic paths. The stochastic path describes which scenario is active at each time step. There can be multiple stochastic paths in parallel. The stochastic structure collects the stochastic paths in a direct acyclic graph (DAG).

But let's make that more clear with an example. We can continue from the previous structure, but let's rename the structure and scenarios. (optional step)

• Right click the object (either in the tree view or the graph view) and select edit
• Rename the stochastic structure from deterministic to DAG
• Rename the realization scenario to base
• Rename the independent scenario to forecast1

Perhaps from the name you already guessed it, we are going to add some scenarios.

• Add two scenario objects forecast2 and forecast3
• Connect the two scenarios to stochastic structure

And we need to adjust the map for the electricity demand accordingly.

• Edit the map and provide a value for each scenario

(see image below)

All these scenarios are independently available to the stochastic structure but now we want to define the underlying relationships to make a stochastic path. In particular, we want to start from a base scenario and later split in the forecast scenarios. For SpineOpt that means that the base scenario is the parent scenario and the following forecast scenarios are the child scenarios.

• add the parent_stochastic_structure__child_stochastic_structure

for each forecast scenario and select the base scenario as its parent (the first scenario is the parent scenario and the second scenario is its child)

We also need to tell SpineOpt what the probability is that we end up in a certain child. That information is stored in the stochastic structure so you'll find the corresponding parameter in the stochastic_structure__stochastic_scenario relationship. Here we assume that each forecast is equally likely to happen.

• for each DAG | forecast relationship, add a value for

the weight_relative_to_parent parameter; the sum needs to be equal to 1

That results in the stochastic structure below.

We can run the SpineOpt tool on this database but we will only see the values for the base scenario. That is because SpineOpt assumes that a scenario runs forever. So, we need to tell SpineOpt when the base scenario ends.

• The current resolution of the system is 1D

but we need a higher resolution if we want to switch scenarios. So, set the resolution parameter of the temporal block flat to 1h.

• To end the base structure after 6 h,

we go to the DAG | base relationship and set the parameter stochastic_scenario_end to a 6h duration value (to obtain a duration value we need to right click the value field and select the parameter type duration)

Do not forget to save/commit from time to time.

When we run the model now, we will obtain values for all scenarios.