Feedback when running in a medium-sized case study

[datejada]

When testing the ModelAnalyzer.jl in a medium-sized case study in Tulipa, we found Numerical analysis found 79309 issues, but we know now what to do 😅 and the package will help us to improve the case study and input data (see TulipaEnergy/Tulipa-OBZ-CaseStudy#38).

During the process, we have the following feedback on the package:

• The verbose keyword is not working as expected. Either true or false throws several lines in the REPL.
• The following commands that are in the docs threw errors:
  • https://github.com/jump-dev/ModelAnalyzer.jl/blob/eb324011051aa3be61e539976e169efdc1700fb8/docs/src/index.md?plain=1#L121 ERROR: MethodError: no method matching summarize(::ModelAnalyzer.Numerical.Data, ::Type{ModelAnalyzer.Numerical.VariableBoundAsConstraint})
  • https://github.com/jump-dev/ModelAnalyzer.jl/blob/eb324011051aa3be61e539976e169efdc1700fb8/docs/src/index.md?plain=1#L127 ERROR: MethodError: no method matching summarize(::ModelAnalyzer.Numerical.Data, ::Vector{ModelAnalyzer.Numerical.VariableBoundAsConstraint})
  • https://github.com/jump-dev/ModelAnalyzer.jl/blob/eb324011051aa3be61e539976e169efdc1700fb8/docs/src/index.md?plain=1#L130 ERROR: MethodError: no method matching summarize(::ModelAnalyzer.Numerical.Data, ::ModelAnalyzer.Numerical.VariableBoundAsConstraint)

This is the full report my_report.txt

Here is the LP file for reproducibility: https://filesender.surf.nl/?s=download&token=3ef84db3-7bfd-4101-a8b4-5af77b4761a2

Julia code:

using JuMP
using ModelAnalyzer
using HiGHS

# Create a JuMP model from lp file from OBZ case study
model = read_from_file("tulipa-obz-model-10rp.lp")
set_optimizer(model, HiGHS.Optimizer)
optimize!(model)

# Perform a numerical analysis of the model
data = ModelAnalyzer.analyze(ModelAnalyzer.Numerical.Analyzer(), model)
# print report
ModelAnalyzer.summarize(data; verbose=false)

Julia environment:

(TestModelAnalyzer) pkg> st
Status `C:\Users\tejadaarangoda\OneDrive - TNO\00_Sandbox\TestModelAnalyzer\Project.toml`
  [87dc4568] HiGHS v1.16.0
  [4076af6c] JuMP v1.25.0
  [d1179b25] ModelAnalyzer v0.1.0 `https://github.com/jump-dev/ModelAnalyzer.jl#main`

I hope this information helps

[joaquimg]

With latest master I get:

julia> ModelAnalyzer.summarize(data; verbose=false)
## Numerical Analysis

## Configuration

  Dense fill-in threshold: 0.1
  Dense entries threshold: 1000
  Small coefficient threshold: 1.0e-5
  Large coefficient threshold: 100000.0

## Dimensions

  Number of variables: 62456
  Number of constraints: 111205
  Number of nonzeros in matrix: 295235
  Constraint types:
    * AffExpr-MathOptInterface.EqualTo{Float64}: 32255
    * AffExpr-MathOptInterface.GreaterThan{Float64}: 25755
    * AffExpr-MathOptInterface.LessThan{Float64}: 53195
    * VariableRef-MathOptInterface.GreaterThan{Float64}: 53116

## Coefficient ranges

  Matrix:    [2e-01, 1e+03]
  Objective: [3e-01, 3e+02]
  Bounds:    [1e+00, 5e+07]
  RHS:       [2e-03, 1e+09]



# VariableBoundAsConstraint
## Number of issues

Found 66590 issues

## List of issues

Showing first 10 issues (66580 issues ommitted)

 * consumer_balance_OBZLL_E_Demand,2050,1,1_1_
 * consumer_balance_OBZLL_E_Demand,2050,1,2_2_
 * consumer_balance_OBZLL_E_Demand,2050,1,3_3_
 * consumer_balance_OBZLL_E_Demand,2050,1,4_4_
 * consumer_balance_OBZLL_E_Demand,2050,1,5_5_
 * consumer_balance_OBZLL_E_Demand,2050,1,6_6_
 * consumer_balance_OBZLL_E_Demand,2050,1,7_7_
 * consumer_balance_OBZLL_E_Demand,2050,1,8_8_
 * consumer_balance_OBZLL_E_Demand,2050,1,9_9_
 * consumer_balance_OBZLL_E_Demand,2050,1,10_10_


# LargeBoundCoefficient
## Number of issues

Found 22 issues

## List of issues

Showing first 10 issues (12 issues ommitted)

 * storage_level_over_clustered_year_NO_Hydro_Reservoir,2050,365_365_ : 5.374644e7
 * storage_level_over_clustered_year_SE_Hydro_Reservoir,2050,365_365_ : 1.5965663e7
 * storage_level_over_clustered_year_ES_Hydro_Reservoir,2050,365_365_ : 5.82425e6
 * storage_level_over_clustered_year_CH_Pump_Hydro_Open,2050,365_365_ : 5.28e6
 * storage_level_over_clustered_year_FR_Hydro_Reservoir,2050,365_365_ : 5.0e6
 * storage_level_over_clustered_year_ES_Pump_Hydro_Open,2050,365_365_ : 2.98095e6
 * storage_level_over_clustered_year_IT_Hydro_Reservoir,2050,365_365_ : 2.783925e6
 * storage_level_over_clustered_year_FI_Hydro_Reservoir,2050,365_365_ : 2.765e6
 * storage_level_over_clustered_year_GR_Hydro_Reservoir,2050,365_365_ : 1.8e6
 * storage_level_over_clustered_year_RO_Hydro_Reservoir,2050,365_365_ : 1.21265e6


# LargeRHSCoefficient
## Number of issues

Found 12697 issues

## List of issues

Showing first 10 issues (12687 issues ommitted)

 * max_output_flows_limit_EU_SMR_CCS,2050,1,1_1_ : 9.99999999e8
 * max_output_flows_limit_EU_SMR_CCS,2050,1,2_2_ : 9.99999999e8
 * max_output_flows_limit_EU_SMR_CCS,2050,1,3_3_ : 9.99999999e8
 * max_output_flows_limit_EU_SMR_CCS,2050,1,4_4_ : 9.99999999e8
 * max_output_flows_limit_EU_SMR_CCS,2050,1,5_5_ : 9.99999999e8
 * max_output_flows_limit_EU_SMR_CCS,2050,1,6_6_ : 9.99999999e8
 * max_output_flows_limit_EU_SMR_CCS,2050,1,7_7_ : 9.99999999e8
 * max_output_flows_limit_EU_SMR_CCS,2050,1,8_8_ : 9.99999999e8
 * max_output_flows_limit_EU_SMR_CCS,2050,1,9_9_ : 9.99999999e8
 * max_output_flows_limit_EU_SMR_CCS,2050,1,10_10_ : 9.99999999e8

Or

julia> ModelAnalyzer.summarize(data; verbose=true)
## Numerical Analysis

## Configuration

  Dense fill-in threshold: 0.1
  Dense entries threshold: 1000
  Small coefficient threshold: 1.0e-5
  Large coefficient threshold: 100000.0

## Dimensions

  Number of variables: 62456
  Number of constraints: 111205
  Number of nonzeros in matrix: 295235
  Constraint types:
    * AffExpr-MathOptInterface.EqualTo{Float64}: 32255
    * AffExpr-MathOptInterface.GreaterThan{Float64}: 25755
    * AffExpr-MathOptInterface.LessThan{Float64}: 53195
    * VariableRef-MathOptInterface.GreaterThan{Float64}: 53116

## Coefficient ranges

  Matrix:    [2e-01, 1e+03]
  Objective: [3e-01, 3e+02]
  Bounds:    [1e+00, 5e+07]
  RHS:       [2e-03, 1e+09]



# `VariableBoundAsConstraint`

## What

A `VariableBoundAsConstraint` issue is identified when a constraint is
equivalent to a variable bound, that is, the constraint has only one
non-zero coefficient, and this coefficient is equal to one.

## Why

This can be a sign of a mistake in the model formulation. Variable
bounds are frequently used by solver in special ways that can lead to
better performance.

## How to fix

Remove the constraint and use the variable bound directly.

## More information

- https://support.gurobi.com/hc/en-us/community/posts/24066470832529/comments/24183896218385

## Number of issues

Found 66590 issues

## List of issues

Showing first 10 issues (66580 issues ommitted)

 * Constraint: consumer_balance_OBZLL_E_Demand,2050,1,1_1_
 * Constraint: consumer_balance_OBZLL_E_Demand,2050,1,2_2_
 * Constraint: consumer_balance_OBZLL_E_Demand,2050,1,3_3_
 * Constraint: consumer_balance_OBZLL_E_Demand,2050,1,4_4_
 * Constraint: consumer_balance_OBZLL_E_Demand,2050,1,5_5_
 * Constraint: consumer_balance_OBZLL_E_Demand,2050,1,6_6_
 * Constraint: consumer_balance_OBZLL_E_Demand,2050,1,7_7_
 * Constraint: consumer_balance_OBZLL_E_Demand,2050,1,8_8_
 * Constraint: consumer_balance_OBZLL_E_Demand,2050,1,9_9_
 * Constraint: consumer_balance_OBZLL_E_Demand,2050,1,10_10_


# `LargeBoundCoefficient`

## What

A `LargeBoundCoefficient` issue is identified when a variable has a
bound with a large absolute value.

## Why

Large bounds can lead to numerical instability in the solution process.

## How to fix

Check if the bound is correct. Check if the units of variables and
coefficients are correct. Check if the number makes is
reasonable given that solver have tolerances. Sometimes these
bounds can be replaced by zeros.

## More information

- https://jump.dev/JuMP.jl/stable/tutorials/getting_started/tolerances/

## Number of issues

Found 22 issues

## List of issues

Showing first 10 issues (12 issues ommitted)

 * Variable: storage_level_over_clustered_year_NO_Hydro_Reservoir,2050,365_365_ with bound 5.374644e7
 * Variable: storage_level_over_clustered_year_SE_Hydro_Reservoir,2050,365_365_ with bound 1.5965663e7
 * Variable: storage_level_over_clustered_year_ES_Hydro_Reservoir,2050,365_365_ with bound 5.82425e6
 * Variable: storage_level_over_clustered_year_CH_Pump_Hydro_Open,2050,365_365_ with bound 5.28e6
 * Variable: storage_level_over_clustered_year_FR_Hydro_Reservoir,2050,365_365_ with bound 5.0e6
 * Variable: storage_level_over_clustered_year_ES_Pump_Hydro_Open,2050,365_365_ with bound 2.98095e6
 * Variable: storage_level_over_clustered_year_IT_Hydro_Reservoir,2050,365_365_ with bound 2.783925e6
 * Variable: storage_level_over_clustered_year_FI_Hydro_Reservoir,2050,365_365_ with bound 2.765e6
 * Variable: storage_level_over_clustered_year_GR_Hydro_Reservoir,2050,365_365_ with bound 1.8e6
 * Variable: storage_level_over_clustered_year_RO_Hydro_Reservoir,2050,365_365_ with bound 1.21265e6


# `LargeRHSCoefficient`

## What

A `LargeRHSCoefficient` issue is identified when a constraint has a
right-hand-side with a large absolute value.

## Why

Large right-hand-sides can lead to numerical instability in the solution
process.

## How to fix

Check if the right-hand-side is correct. Check if the units of variables
and coefficients are correct. Check if the number makes is
reasonable given that solver have tolerances. Sometimes these
right-hand-sides can be replaced by zeros.

## More information

- https://jump.dev/JuMP.jl/stable/tutorials/getting_started/tolerances/

## Number of issues

Found 12697 issues

## List of issues

Showing first 10 issues (12687 issues ommitted)

 * Constraint: max_output_flows_limit_EU_SMR_CCS,2050,1,1_1_ with right-hand-side 9.99999999e8
 * Constraint: max_output_flows_limit_EU_SMR_CCS,2050,1,2_2_ with right-hand-side 9.99999999e8
 * Constraint: max_output_flows_limit_EU_SMR_CCS,2050,1,3_3_ with right-hand-side 9.99999999e8
 * Constraint: max_output_flows_limit_EU_SMR_CCS,2050,1,4_4_ with right-hand-side 9.99999999e8
 * Constraint: max_output_flows_limit_EU_SMR_CCS,2050,1,5_5_ with right-hand-side 9.99999999e8
 * Constraint: max_output_flows_limit_EU_SMR_CCS,2050,1,6_6_ with right-hand-side 9.99999999e8
 * Constraint: max_output_flows_limit_EU_SMR_CCS,2050,1,7_7_ with right-hand-side 9.99999999e8
 * Constraint: max_output_flows_limit_EU_SMR_CCS,2050,1,8_8_ with right-hand-side 9.99999999e8
 * Constraint: max_output_flows_limit_EU_SMR_CCS,2050,1,9_9_ with right-hand-side 9.99999999e8
 * Constraint: max_output_flows_limit_EU_SMR_CCS,2050,1,10_10_ with right-hand-side 9.99999999e8

[joaquimg]

The advanced methods were off indeed.
Just fixed methods and docs in #27

[joaquimg]

Closing as solved