Run settings

Run Settings can be found in the Views group of the Home ribbon tab.

Default values for run settings are shown in light gray. When you enter a different value, this is shown in black. If you delete the value you entered, the default value is restored.

Note: When you change run settings for a case in your model, the change cascades to all descendants cases, unless the particular setting has been overridden. If the setting has been overridden in a subcase, the override is preserved for the subcase and all its children cases, and changes from parent cases are not inherited.
Any subcases that have been copied and pasted from another model do not inherit changes in run settings from parent cases.
Once the inheritance chain for run settings is broken, it cannot be restored. You can still change run settings for multiple cases at once by opening them as a case stack.
For more information about the case hierarchy in a model, see this topic.

Infeasibility breakers

If your model does not optimize, you can use infeasibility breakers to allow it to violate one or more constraints. This will help you understand the source of the infeasibility, by finding a convergent solution. Each constraint that is allowed to be violated in the optimization has an associated infeasibility breaker penalty. Violating the constraint at the solution will activate the penalty and impact the optimizer objective function.

When infeasibility breakers are activated in an optimization, the solution metrics report the associated penalty cost and the number of activated infeasibility breakers. In addition, the Last Run Diagnostics window provides error diagnostics for each violated constraint, along with the penalty cost associated with the violation.

The Infeasibility Breaker Settings page lets you configure the types of constraints that can be violated and the penalty costs associated with breaking them.

From the Activity drop-down list you can choose when to enable infeasibility breakers: Always, Never or Only When Infeasible. In the last case, a first optimization run without infeasibility breakers is attempted. If the problem is infeasible, it is optimized again with infeasibility breakers active.

You can enable or disable whole categories of infeasibility breakers by selecting or clearing the check boxes in the Infeasibility Breaker Settings page. You can also define the penalties costs associated with breaking the constraints.

You can also enable or disable individual infeasibility breakers in other locations of the planning environment, such as the Purchases and Sales pages. Since the infeasibility breaker column may be hidden by default on these pages, see this topic on how to change which columns are visible.

Optimization starting point

The Optimization Starting Point settings are used to improve the starting conditions for optimization.

There are three settings you can enable.

Enable Multi-Start. Select this option to use multiple start points.

When you enable multi-start, enter the number of start points you want to use. A sensible initial number is 4-6 start points, but more may be necessary.

After the multi-start optimization has finished, the Multi-Start Metrics and Multi-Start Metrics Chart report the solution information for each start point.

Enter a percentage value in the Partial Completion field if you want to finish the optimization when a given fraction of the multi-start points have completed. For example, with ten start points and a completion percentage of 80%, the optimization finishes when eight points have completed, without waiting for the last two points. Fractional numbers of points are rounded up: for example, a completion percentage of 50% with 15 starting points results in the optimization ending when eight points have completed (15 × 0.5 = 7.5, rounded up to 8).

Note: When enough points have completed to satisfy the partial completion percentage, the optimizer runs for another 20% of the elapsed time to allow further points to complete. For example, if enough points have completed after 20 seconds, the optimizer runs for another 20 × 0.2 = 4 seconds.

The aim of the Partial Completion feature is to stop processing any start points that do not converge at all or take significantly longer to converge, often to a local optimum which does not improve the overall objective function. By using this feature, you can increase the number of converged start points while decreasing the optimization time, because the optimization does not have to wait for all the points to converge or time out.

Example: You may use eight start points for your model, with partial completion disabled. Six of these points converge, while two time out, greatly increasing the overall optimization time.
You then choose to run 12 points with a 66% partial completion percentage. You then obtain at least eight converged points, while excluding the two failing points.

The optimal partial completion percentage value depends on the model. Reasonable initial values range from 66% to 75%, potentially paired with an increase in the number of start points.

Note: The Partial Completion option does not work with the Calculation Hub when multiple start points are being run across multiple nodes.

Enable Warm Start. Select this option to use the solution from the selected case as the starting point of the optimization.

Enable Randomization Control. Select this option to customize which quantities in the case are randomized before optimization. This gives you granular control over the initial starting point of the optimization.

You can enable randomization for four categories:

• Flow Routing, including initial split fractions for splitters, blenders and other instances where flows are split (for example among different modes of a process unit).

• Routing for Inventories of crudes, imports and products.

• Parameters, including operating and global parameters.

• Purchases of crudes, imports and utilities.

If you disable randomization control for a category, the starting points for the affected quantities are set as follows:

• If Enable Warm Start is selected, the values in the selected case are used as starting points.

• Otherwise, the starting points are set deterministically, as they would for a single start point without multi-start.

  Warning: If you select Enable Randomization Control but leave all the four categories unselected, the setting has no effect and no randomization of the starting point occurs.
  If Enable Multi-Start is selected in this situation, it means that the same starting point is used every time. This results in an optimization error.

  See the following table for the effect of various combinations of settings on this page. The warning () and error () icons in the last column indicate combinations that result in diagnostic warning or error messages.

  Multi-start	Warm start	Randomization control	Randomization control items	Resulting behavior
  Disabled	Disabled	Disabled	None selected	Standard single start optimization.
  Enabled	Disabled	Disabled	None selected	Standard multi start optimization.
  Disabled	Enabled	Disabled	None selected	Standard warm start optimization.
  Enabled	Enabled	Disabled	None selected	Runs the same warm start optimization multiple times.
  Disabled	Disabled	Enabled	None selected	Standard single start optimization. The randomization control setting has no effect because no items are selected.
  Enabled	Disabled	Enabled	None selected	Runs the same non-randomized optimization multiple times. The randomization control setting has no effect because no items are selected.
  Disabled	Enabled	Enabled	None selected	Standard warm start optimization. The randomization control setting has no effect because no items are selected.
  Enabled	Enabled	Enabled	None selected	Runs the same warm start optimization multiple times. The randomization control setting has no effect because no items are selected.
  Disabled	Disabled	Enabled	One or more selected	Standard single start optimization. The randomization control settings have no effect.
  Enabled	Disabled	Enabled	One or more selected	Multi-start optimization with one or more randomized items.
  Disabled	Enabled	Enabled	One or more selected	Warm start optimization with a single start point and one or more randomized items.
  Enabled	Enabled	Enabled	One or more selected	Warm start optimization with multiple start points and one or more randomized items.

Result retrieval settings

When an option in this list is selected, results of the corresponding type are retrieved and shown in the user interface. Retrieving all results, especially marginal values, may slow down result display and report generation.

Xpress-SLP settings

The Xpress-SLP Settings control the behavior of the optimizer used in Plan. These settings are only relevant when the run type is set to Optimize.

Note: There should be no need for you to change any of these settings without advice from AVEVA.

Maximum SLP iterations and maximum LP iterations

The optimizer uses sequential linear programming which iteratively solves a series of linear problems to find the optimal solution.

The maximum number of iterations used to solve each linear problem is controlled by Maximum LP iterations.

The maximum number of times a new linear approximation is built and solved is controlled by Maximum SLP Iterations.

The maximum run time (in seconds) that the optimizer runs for when trying to find a solution is controlled by Maximum SLP Run Time.

If Activity in the Infeasibility Breaker Setting page is set to Only When Infeasible, the effective maximum SLP run time is twice the value in this field:

1. The optimization runs to the maximum time to find a solution.

2. If the optimization times out because the model is infeasible, the infeasibility breaker setting is invoked. The optimization runs again with breakers turned on, which either leads to a solution or to another time out. In this case, the Max Start Point Run Time setting in the Advanced Settings page covers both optimization runs.

Cascading active

The linear approximations built by the solver can be improved in quality by evaluating non-linear terms in a specific order, this is called Cascading by FICO®, the developers of the optimizer used in Plan.

This is an optimizer-specific setting for the FICO® optimizer.

Solver algorithm

The optimizer default setting is to employ the Simplex method, but in some cases the Newton barrier interior-point method may solve the model more quickly and robustly. To test the performance of your model using the barrier method, select Barrier in the Solver Algorithm list. The algorithm takes a number of Barrier Passes (the default is 10) and has an option to Enable Crossover.

For more information on the barrier algorithm and its parameters, please consult the FICO® Xpress documentation.

Advanced settings

The Advanced Settings control some of the advanced behavior of the optimization problem, along with the manner in which results are retrieved from the optimized solution.

Max Start Point Run Time. This setting allows you to define an upper limit in seconds for each starting point in a multi-start optimization. A start point optimization is canceled if it reaches the maximum start point run time. Note that the Max Start Point Run Time value must be at least equal to the Maximum SLP Run Time setting in the Xpress-SLP Settings page. If not, it is increased automatically during optimization.

Dynamic bounding helps deal with scenarios where there is zero flow in streams. Calculating property values for zero flows is not possible, making it difficult during optimization to determine the impact of increasing the flow to the stream. Dynamic bounding uses constraints to ensure that flow rates are not zero.

Partial Relaxation in Early Iterations. If you select this option, dynamic bounding constraints are dropped after the first and subsequent iterations for sufficiently non-zero flows.

Downstream Stream Values Contributions. If you select this option, an extra unit of material is always forced to return downstream contributions to a stream value, irrespective of downstream constraints preventing the flow of additional material. Large negative values, shown as cells with a blue background in the Stream Values page, are then associated with the cut yields, properties or calculations responsible for the material being rejected.

When this option is not selected, the same downstream constraints force the optimizer to buy one fewer unit of material upstream, because the extra unit of material in the stream must displace a unit of original material. Downstream contributions are not available because in this case the value comes from upstream, as one fewer material unit purchased.

Note: There should be no need for you to change any of these settings, apart from Max Start Point Run Time and the recommended settings described below, without specific instructions from AVEVA Support.

Restore settings

Click the icon to open the Restore Run Settings dialog box, with the following options:

• Choose Use default settings to restore the run settings to their original default values.

• Choose Use parent case settings to restore the run settings to the values of the parent case in the model hierarchy. This option is disabled in two cases:

  • If you have opened the master case, which does not have a parent.

  • If all the values are already inherited from the parent case, with no overrides. Note that typing a value in a field always counts as an override, even if the value is the same as the inherited one.