Multi-period models

Multi-period modeling allows you to take into account future operations and market opportunities when optimizing plant operations. Multi-period operations help stabilize production where there is uncertainty in pricing and demand, and where future plant activities require changes in plant operations.

Multi-period modeling uses different cases to model each of the periods. Each period can have different pricing, constraints, stream routings, products and so on. Periods are linked by transferring inventory, with the closing inventory in a period immediately available for use at the opening of the subsequent period.

Two workflows are available to optimize multi-period models:

• Sequential roll-forward optimization

• Dynamic classical optimization

Multi-period inventory

Inventory in multi-period modeling is used to allow the results of one period affect the subsequent periods. Inventory is then a method of communication between periods. For example, it may be optimal to stockpile inventory for a future period in order to offset a problem or to take advantage of price differentials.

Follow these steps to configure inventory for multi-period models.

1. Add the necessary tanks to hold inventory

   Inventory is built up and drawn down from tanks. Configure and connect the necessary tanks in your flowsheet.

   Warning: Tank modes
   Where a tank has a mode, the inventory can only be associated with one mode of the tank.
   Pooled inventory
   A tank may be downstream of a process unit whose outputs are pooled (Model Connections & Flowsheet Pipes). As such, the inventory will consist of several separate subpools of material (one for each substream). When routing the inventory to its final destination, it may be necessary to depool the inventory components and route these to appropriate process models (Input Stream Associations).

2. Create planning inventories

   Create a planning inventory for each inventory that should be built up or drawn from.

   Tip: During multi-period optimization, the inventory amount and quality are optimized. Amount is optimized by choosing the amount to hold back versus the amount to send for downstream processing, whereas quality is optimized by controlling the ratio of materials going into inventory (if there is more than one material that can constitute the inventory) and the conditions of the upstream processes which form the inventory (for example the process unit operating parameters and feedstocks).
   You can associate a grade with an inventory. This constrains the qualities of the inventory within the allowed grade specifications. The grade must exist in the reference data, but does not necessarily have to be a finished product grade.
   For example, you could have a final sales naphtha grade and a reduced spec naphtha grade, with slightly looser quality specifications. Within a model you may have an inventory for naphtha prior to the reformer to allow for shutdown optimization. The inventory would be associated with the reduced spec grade, while you also had a sale for the final naphtha grade. During a multi-period optimization the inventory would always have to meet the reduced spec grade, which would ensure that operationally the naphtha itself had reasonable usable qualities, whilst there would also be the final sales grade so that if economic selling naphtha would also be an option.

3. Associate the planning inventory with relevant tanks

   For each planning inventory select the tanks where the inventory can be built up.

   Warning: Storing intermediate distillation products in an inventory between periods is not supported. An intermediate distillation product is an output of a distillation unit and a feed to another, such as atmospheric residue.
   If you need to create a model with this topology, please contact AVEVA Support (spiral.support@aveva.com) to discuss a suitable workaround.

4. Create per-period cases

   Create the cases for each period. In each case ensure that the inventory is active. If necessary, enter constraints on the minimum or maximum inventory amounts per period.

   When configuring the per-period cases, ensure the planning period is correctly set.

5. Create and run the multi-period model

   Convert the model to a multi-period model and run the optimization (for sequential multi-period models or classical multi-period models).

Multi-period constraints

In sequential multi-period optimization, each period is in effect independent, although latter periods depend on prior periods. As each period is independent, it is difficult to define cross-period constraints and there is no automatic way to ensure that cross-period constraints are met. Instead, any cross-period constraint must be managed on a per-period basis according to the business requirements.

It may be possible to use global parameters to help manage cross-period constraints in sequential models. For example, to limit the cumulative use of catalyst across several periods you could define a global parameter to count the amount of catalyst used per period, and sum this across time. However, the initial periods may use all the catalyst allocation, leaving no catalyst for subsequent periods. In this case it would be necessary to limit the catalyst usage manually for each period.

In classical multi-period optimization, you can define cross-period constraints as calculations across the entire set of periods or across a subset.