FEATURE: Robust partition evaluation (#17)

* ADD: robust partition evaluation

* FIX: remove extraneous ONM. function calls

* FIX: wrong test file path

* ADD: Random to extra deps

* FIX: function arg types

* ADD: settings to case

* FIX: input data

* Re-enable tests

CHANGELOG.md

@@ -4,6 +4,10 @@
 
 - none
 
+## v3.6.0
+
+- Added capability to evaluate the optimality of a given partition against a set of load scenarios to judge robustness
+
 ## v3.5.1
 
 - Fixed bug in `constraint_mc_power_balance_shed_block` (`LPUBFDiagPowerModel` version) where `pd_zblock_zdemand` was used instead of `qd_zblock_zdemand`

Project.toml

@@ -1,7 +1,7 @@
 name = "PowerModelsONM"
 uuid = "25264005-a304-4053-a338-565045d392ac"
 authors = ["David M Fobes <[email protected]>"]
-version = "3.5.1"
+version = "3.6.0"
 
 [deps]
 ArgParse = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
@@ -60,7 +60,8 @@ Ipopt = "b6b21f68-93f8-5de0-b562-5493be1d77c9"
 JSON = "682c06a0-de6a-54ab-a142-c8b1cf79cde6"
 Juniper = "2ddba703-00a4-53a7-87a5-e8b9971dde84"
 PowerModelsDistribution = "d7431456-977f-11e9-2de3-97ff7677985e"
+Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [targets]
-test = ["Test", "HiGHS", "Ipopt", "JSON", "Juniper", "PowerModelsDistribution"]
+test = ["Test", "HiGHS", "Ipopt", "JSON", "Juniper", "PowerModelsDistribution", "Random"]

src/PowerModelsONM.jl

@@ -114,6 +114,7 @@ module PowerModelsONM
     include("prob/mld_block.jl")
     include("prob/partitions.jl")
     include("prob/mld_block_robust.jl")
+    include("prob/robust_evaluation.jl")
     include("prob/stability.jl")
     include("prob/switch.jl")
 

src/prob/robust_evaluation.jl

@@ -0,0 +1,63 @@
+"""
+    evaluate_partition_optimality(
+        data,
+        load_scenarios,
+        model_type,
+        solver;
+        save_partial_results,
+        partial_result_folder,
+        time_elapsed,
+        kwargs...
+    )
+
+Function to evaluate the optimality of a specific partition by considering a collection of load scenarios.
+
+`data` has the partition configuration applied.
+"""
+function evaluate_partition_optimality(
+    data::Dict{String,<:Any},
+    load_scenarios,
+    model_type::Type,
+    solver;
+    save_partial_results::Bool=false,
+    partial_result_folder::String=".",
+    time_elapsed::Union{Missing,Real} = missing,
+    kwargs...)
+
+    _results = Dict{String,Any}()
+
+    for ls in keys(load_scenarios)
+        single_load_scenario = Dict{String,Dict{String,Any}}()
+        single_load_scenario["1"] = load_scenarios[ls]
+
+        eng = deepcopy(data)
+
+        if !ismissing(time_elapsed)
+            eng["time_elapsed"] = time_elapsed
+        end
+
+        @debug "starting load scenario evaluation $(ls)/$(length(load_scenarios))"
+
+        result = solve_robust_block_mld(eng, model_type, solver, single_load_scenario; kwargs...)
+
+        if save_partial_results
+            open("$(partial_result_folder)/result_$(ls).json", "w") do io
+                JSON.print(io, result)
+            end
+        end
+
+        _results[ls] = result
+    end
+
+    return _results
+end
+
+
+"""
+retrieve_load_scenario_optimality(results::Dict)
+
+Returns a Dict of objectives for the different load scenarios considered in the robust partition evaluation.
+"""
+function retrieve_load_scenario_optimality(results::Dict{String,<:Any})::Dict{String,Real}
+    return Dict{String,Real}("$i" => results["$i"]["1"]["objective"] for i in 1:length(results))
+end

test/robust_eval.jl

@@ -0,0 +1,47 @@
+function randomize_partition_config(
+    case,
+    num_closed_switches)
+
+    if num_closed_switches > length(keys(case["switch"]))
+        error("Number of closed switches exceeds the total number of switches")
+    end
+
+    part_config = Dict{String,Any}()
+
+    switch_keys = collect(keys(case["switch"]))
+    shuffled_keys = Random.shuffle(switch_keys)
+    closed_switches = shuffled_keys[1:num_closed_switches]
+
+    # Set the status of the selected switches to "CLOSED" and the rest to "OPEN"
+    for key in switch_keys
+        if key in closed_switches
+            part_config[key] = PMD.SwitchState(1)
+        else
+            part_config[key] = PMD.SwitchState(0)
+        end
+    end
+
+    case["fixed_partition_config"] = part_config
+
+    return case
+end
+
+
+@testset "robust partition evaluation" begin
+    case = parse_file("../test/data/ieee13_feeder.dss")
+    settings = parse_settings("../test/data/ieee13_settings.json")
+    PMD.apply_voltage_bounds!(case)
+    case = randomize_partition_config(case, 2)
+
+    num_load_scenarios = 20
+    uncertainty_val = 0.2
+    ls = generate_load_scenarios(case, num_load_scenarios, uncertainty_val)
+
+    solver = optimizer_with_attributes(HiGHS.Optimizer, "primal_feasibility_tolerance" => 1e-6, "dual_feasibility_tolerance" => 1e-6, "small_matrix_value" => 1e-12, "allow_unbounded_or_infeasible" => true, "output_flag" => false)
+
+    results_eval_optimality = evaluate_partition_optimality(case, ls, PMD.LPUBFDiagPowerModel, solver)
+
+    optimality = retrieve_load_scenario_optimality(results_eval_optimality)
+
+    @test isapprox(optimality["1"], 5, atol=1e0)
+end

test/runtests.jl

@@ -15,6 +15,9 @@ import Juniper
 import Ipopt
 import HiGHS
 
+import Random
+Random.seed!(21)
+
 minlp_solver = optimizer_with_attributes(
     Juniper.Optimizer,
     "nl_solver" => optimizer_with_attributes(Ipopt.Optimizer, "tol"=>1e-6, "print_level"=>0),
@@ -60,6 +63,8 @@ silence!()
     # problems
     @info "Running tests in mld.jl"
     include("mld.jl")
+    @info "Running tests in robust_eval.jl"
+    include("robust_eval.jl")
     @info "Running tests in nlp.jl"
     include("nlp.jl")
     @info "Running tests in opf.jl"