GA Model Definition tab

The drop-down list contains the available goal functions for optimization. An additional input field will be displayed where you can specify a comparison value if the function requires it. The goal functions are: Minimize (the algorithm will search for the smallest value of the optimization indicator), Maximize (the algorithm will search for the largest value of the optimization indicator), Closest to (the algorithm will search for the value closest to the specified number; the closest value found may be greater or less than the specified number), Minimal greater than (the algorithm will search for the minimal value of the optimization indicator which is greater than the specified number), Maximal less than (the algorithm will search for the maximal value of the optimization indicator which is less than the specified number).

The settings in this table tell the algorithm which actions can be performed to produce the desired results; all of the options are described in the table below. Note that the more parameters you use, the bigger the search space will be, which will make optimization more likely to fail unless the parameters are selected very carefully. Also make sure that all of the selected parameters are realistic: for example, don't select the Change WI option if projects cannot have equity changes.

Note: These options take project types into account project type settings (see Project Types). For example, if a project belongs to a non-divestible type, it will not be excluded even if the Include and Exclude Projects box is checked.

This table determines which periodic variable targets will be considered during optimization. The targets are set up in the configuration settings (see Periodic Variables) and their values are entered on the Set Targets tab (see Targets). The Target Type column is populated automatically and cannot be edited.

To include a target into optimization, select the Hard or Soft option in the Constraint Type column. If a generated portfolio does not meet a hard target, then it will be discarded. However, if it does not meet a soft target, then a penalty function will be applied. Penalty functions and their variations are described in the table below. For more information.

It is recommended to gradually tighten the set of constraints. For example, start with a single soft constraint on production, find out how well this target is met, then identify a viable target and switch it to a hard constraint, or add additional constraints such as capex. Also note that the GA algorithm does not have to start from the initial random portfolio; it can be start from where a previous run has finished, and thus it can start from a previously-reached goal and search beyond it.

This box is always checked as at least one kind of change to project settings is needed for optimization.

Note: If a MEG is included in optimization and the Include and Exclude Projects box is unchecked, only the previously included project from that MEG will be considered for optimization. If this box is checked, a different project from that MEG may be included. For more information, see Mutually Exclusive Groups.