PTDF and LODF based load flow, slack penalties, group balance - [merged] #236

Closed spine-o-bot closed 3 years ago

In GitLab by @DillonJ on Apr 10, 2020, 12:46

_Merges jd_loadflow -> dev

Introduces PTDF based load flow and security constrained unit commitment based on LODFs
PTDFs are calculated in util/network_functions.jl
Area based demand is now based on node groups. The parameter fractional_demand (node_group__node) controls the fraction of a node group's demand which is present at each member node
- Area based balance is enabled using the balance_type parameter. When set to balance_type=balance_type_group, the constraint_group_balance constraint is enabled in place of constraint_nodal_balance. For all members of a node_group for which balance_type=balance_type_group, the constraint_nodal_balance is supressed.
- To enable ptdf based load flow for a commodity, set commodity_physics= commodity_physics_ptdf. To enable lodf based security constrained unit commitment (N-1 contingency constraints) commodity_physics= commodity_physics_lodf. In both cases, balance_type should be set to balance_type_group for the node_groups corresponding to the nodes in question
- node_slack_pos and node_slack_neg variables are activated for a node if node_slack_penalty is set for that node. This is then penalised in the objective function (objective_penalties.jl)

A test input db for this merge is available here:

https://drive.google.com/open?id=1PtCgokfFop9vfbV6HeYCQE4zKxQ9hWVr

In GitLab by @DillonJ on Apr 14, 2020, 06:41

changed the description

In GitLab by @DillonJ on Apr 14, 2020, 06:42

changed the description

In GitLab by @DillonJ on Apr 14, 2020, 07:15

added 1 commit

fa727138 - Removed reference to Area object in Update network_functions.jl

Compare with previous version

In GitLab by @manuelma on Apr 15, 2020, 16:48

I suggest if commodity_physics(commodity=c) in (:commodity_physics_lodf, :commodity_physics_ptdf).

Also, please note that we don't put spaces around the equal sign when used to pass keyword arguments.

In GitLab by @manuelma on Apr 15, 2020, 16:54

This doesn't feel super efficient, I wonder if there's a better way to exclude nodes from the nodal balance...

In GitLab by @manuelma on Apr 15, 2020, 16:56

We avoid globals for both style and performance. Is it difficult to get rid of them?

In GitLab by @manuelma on Apr 15, 2020, 16:58

In general a see issues with style in this file. Please note that we put spaces after commas and around conditional operators, and also around the assignment operator.

In GitLab by @manuelma on Apr 15, 2020, 17:01

You always need to break after the first iteration? I suggest using c = first(indices(commodity_lodf_tolerance)) instead of a for loop.

In GitLab by @manuelma on Apr 15, 2020, 17:02

I suggest putting this in a function so run_spinemodel() stays as lean as possible. I'd like to be able to read along run_spinemodel rather quickly to get what's going on at a high level.

In GitLab by @DillonJ on Apr 15, 2020, 17:12

I agree that it's not particularly nice having this code at the start, but it's the most efficient way I could think of determining if a node is not in a node group with a certain parameter set, since it only has to check the node groups if it's in one to begin with...

What I wanted to avoid was having to set a parameter for each node which would be laborious for a user and would be depending on default values which I also wanted to avoid.

I am open to suggestion!

In GitLab by @DillonJ on Apr 15, 2020, 17:13

I can look into it, but it seemed awkward to me

In GitLab by @DillonJ on Apr 15, 2020, 17:13

Nice, this is the syntax I was looking for