[ANT-4033] Feat: Support setObjectiveOffset

[JasonMarechal25]

Specification: objectiveOffset Calculation

Overview

The objectiveOffset represents the constant term in a linear programming objective function. For an objective function of the form Σ(aᵢ·xᵢ) + b, where xᵢ are decision variables with coefficients aᵢ, the offset corresponds to the constant term b.

This constant term aggregates all non-variable contributions to the objective value and is computed during the model building phase in ComponentFiller::addObjectives().

Calculation Rules

1. Variables Alone

No offset contribution: Variables and their coefficients do not contribute to the offset, regardless of their time/scenario dependency.

• x → offset = 0
• param × x → offset = 0
• x(t) over n time steps → offset = 0

2. Literal Constants

Offset = constant value

• x + 10 → offset = 10

3. Parameters as Constant Terms

Offset = parameter value evaluated according to its time/scenario dependency. Parameters represent known values (not optimization variables) and thus contribute to the constant term.

4. Time Dependency

Variable with constant offset

Offset = n × constant_value (where n is the number of time steps)

• Variable over 10 time steps with constant offset 10 → total offset = 100

Time-varying parameter

Offset = Σₜ param(t) (sum over all time steps in the optimization horizon)

• param = [10, 11, 12] over 3 time steps → offset = 33

5. Scenario Dependency

Offset = value for the selected scenario

Parameters varying only by scenario contribute their value from the selected scenario's time series.

• Parameter with 3 scenarios [15, 25, 35], scenario 2 selected → offset = 35

6. Time AND Scenario Dependency

Offset = Σₜ param(t, s) (sum over all time steps for the selected scenario)

• Parameter in selected scenario with values [15, 25, 35] over 3 time steps → offset = 75

7. Multiple Scenario Groups

When different model elements use different scenario groups:

• Each element selects its time series from its own scenario group independently
• Offsets from different elements accumulate
• Only parameters (constant terms) contribute to the offset
• Variables never contribute to the offset, even when using different scenario groups

Example:

• Variable x using scenarioX with values [11, 22, 33]
• Parameter using scenarioY with values [3, 6, 9] over 3 time steps
• Variable contribution: 0 (variables don't affect offset)
• Parameter contribution: 3 + 6 + 9 = 18
• Total offset: 18

[guilpier-code]

Only time independent expressions are accepted in objectives.
To make it even more clear in the code, we could have, inside the loop over objectives :

const auto root_node = objective.expression().RootNode();
const auto variability = getVariability(root_node, component_);
if (isTimeDependent(variability))
{
    throw Error::RuntimeError("Time dependent objectives are not supported in Antares.");
}

const auto linearExpression = visitor.visitMergeDuplicates(root_node)[0];
addStaticObjective(linearExpression);
objectiveOffset += linearExpression.constant();

Note that, in code above :

• linearExpression is now of type LinearExpression (and no longer of type TimeDependentLinearExpression), so it has 1 dimension (not multiple dimensions).
• addStaticObjective(...) takes a LinearExpression

[JasonMarechal25]

indeed better : d6e0b98 (#3207)

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
73.7% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud