Investigate the use of dictionaries instead of vectors for multi-conductor quantities

[matteorossini]

The current implementation of PowerModelsMCDC uses an hybrid approach to data structures:

• dictionaries as a general data structure, providing hierarchy;
• vectors (only at the leaves of the dict tree) to host multi-conductor quantities.

Problem

If we look at the $k$-th element of a vector, its meaning depends on another variable.

Example

Let's look at converter results in a solution dictionary. The meaning of pdc vector's elements depends on 2 parameters in input data (conv_confi and connect_at).

Bipolar converter:

julia> conv["1"]["pdc"]
3-element Vector{Float64}:
 1.2 # positive pole
 0.1 # metallic return
 1.4 # negative pole

Monopolar converter connected to positive and neutral terminal:

julia> conv["2"]["pdc"]
1-element Vector{Float64}:
 1.2 # positive pole
 0.1 # metallic return

Monopolar converter connected to negative and neutral terminal:

julia> conv["3"]["pdc"]
2-element Vector{Float64}:
 1.2 # negative pole
 0.1 # metallic return

Depending on the converter configuration:

• the first element can refer to either positive or negative pole;
• there may or may not be a third element.

Consequences of the problem

• Developer side: we have to use look-up tables for bookkeeping.
• User side: to interpret the content of solution dict, the input data dict is needed, too.

Why we have this problem?

Vectors have contiguous indices. E.g., cannot have v[2] if there is no v[1].

Possible solution

Use dictionaries.

Bipolar converter:

julia> conv["1"]["pdc"]
Dict{String, Any} with 3 entries:
  "p" => 1.2
  "r" => 0.1
  "n" => 1.4

Monopolar converter connected to positive and neutral terminal:

julia> conv["2"]["pdc"]
Dict{String, Any} with 2 entries:
  "p" => 1.2
  "r" => 0.1

Monopolar converter connected to negative and neutral terminal:

julia> conv["3"]["pdc"]
Dict{String, Any} with 2 entries:
  "r" => 0.1
  "n" => 1.4

Caveats

• Need to investigate whether this choice would lead to unexpected consequences or awkward code patterns.