-
Notifications
You must be signed in to change notification settings - Fork 0
cep25
| CEP: | 25 |
|---|---|
| Title: | Preference Adjustment Process |
| Last-Modified: | 2015-08-25 |
| Author: | Matthew Gidden |
| Status: | Accepted |
| Type: | Standards Track |
| Created: | 2015-07-07 |
This CEP describes the process of adjusting preferences in the DRE, e.g., with regard to implementing international trade instruments. It generalizes methodology of the preference adjustment phase.
The current preference adjustment phase only includes the requesting facility and its managers (i.e. institution and region). Only allowing requesting managers to adjust preferences is problematic because it prohibits, for example, the complete modeling of intra-regional instruments such as tariffs.
In order for managers of bidding regions to apply tariffs, or any other inter-entity instruments, they must be involved in preference adjustment. Previously, only managers of the requesting entities were allowed to affect preference values. This CEP will allow bidders and managers of bidding entities to affect preferences as well.
There is not a clear reason or motivation as-of-yet to allow requesting managers
to adjust preferences before bidding managers (and vice-versa). Accordingly, a
general approach is taken for the preference adjustment ordering
methodology. The entities involved in any trade preference adjustment are the
two traders, and each of their associated managers. In a common Cyclus use case,
this involves two Trader entities whose immediate "manager" is the
Facility entity of whom they are a member. Further, each Facility is
managed by an Institution which is managed by a Region.
For the purpose of this CEP, a default ordering is defined: the requester and its managers are queried in order followed by the managers of the bidder. The bidder is not involved, because it is assumed that suppliers are motivated to reduce inventory. Bidder's managers are involved in order to generally model international instruments, such as tariffs. As different orderings are desired in the future, they may be implemented. The only requirement of a given ordering implementation is that it must maintain a preference adjustment history. That is, a data structure comprising and ordered list of entity-preference pairs that records the history of the adjusted preference for each trade.
An important caveat includes the adjustment of preferences to a "no trade" value. Archetype developers are allowed to denote a trade as "impossible" by setting the preference value to a non-positive number. If at any point during the adjustment phase such a state is found, the phase is considered over for that trade and the trade is removed from the exchange. Consider the following situation: a trade is initially set to the default preference of unity; the first manager queried has a rule that negates the trade (sending its value to negative unity) while the next manager queried has a rule that increases all trades by a magnitude of two units. Rather than setting the final trade preference at unity based on input from both managers, the trade is removed from the exchange after its preference is adjusted by the first manager.
The determination of the final preference value given the adjustment history, i.e., the preference aggregation rule, is also tuneable. One could imagine many different negotiation models, or negotiation rules, by which to determine a final value given a collection of entities' input. Again, this CEP provides a default rule, which is to take the final preference value, unperturbed, through the adjustment process. However, any rule that takes as an argument the adjustment history data structure and provides a final preference value is valid in this methodology. Of course, different rules may be implemented as they are required.
This CEP does not explicitly provide an user-facing interface for adjusting either the adjustment ordering or aggregation. The provided methodology is a novel first step in fuel cycle simulation, and it is not at all clear that full user control over the adjustment process is necessary for the current set of modeled scenarios. If that capability is identified as critical path in the future, a user-facing interface can be implemented that allows adjustment of both the preference adjustment ordering and preference aggregation determination.
See [1] for implementation of Agent-based preference adjustment.
No backwards incompatibilities are introduced with this CEP.
This document is released under the CC-BY 3.0 license.
| [1] | https://github.com/cyclus/cyclus/pull/1122 |