Traditional Commitment Logic

The unit commitment logic in Aurora simulates operation of generating units that cannot cycle on and off hourly. Any unit with a Non-cycling input not equal to zero is defined to be a commitment unit. Depending on the commitment method chosen, these units commit to operate for a specific period based either upon the value they are forecasted to create over the period, or their contribution to system reliability. Once committed, units will run at minimum capacity or above, depending on the value created in each hour of operation. There are two options for the commitment method used in Aurora, self-commitment and pool based commitment. These methods may be used simultaneously in a study, although all units within a zone must generally use the same method. The following sections describe the two methods.

Decision Method 1 – Self-Commitment

This is the default methodology for unit commitment in Aurora and uses a price-based, self-commitment algorithm. This methodology is used for all units which do not belong to an operating pool defined to use the Pool-based Commitment logic described in the next section. All non-cycling units using the self-commitment method will have commitment decisions evaluated and updated for every hour of the dispatch. This method uses zone-specific, 168 hour-ahead, internal market price forecasts to evaluate the economics of unit commit and de-commit decisions. The internal zone forecasts use observed zonal price history in conjunction with other observed simulation parameters to produce the 168 hour-ahead forecast. The internal forecasts are updated dynamically each hour as model chronology proceeds.

At the beginning of each dispatch hour, all non-cycling units are classified according to their commitment eligibility. Units that have been offline for at least their Minimum Down Time are eligible for commitment. Those that have been running for at least their Minimum Up Time are eligible for de-commitment. For commit eligible units, an algorithm is run to determine the unit’s expected pattern of operation and resulting cash flow over its minimum up time, if started in the hour, and compensated according to the hourly price profile contained in the internal forecast. Unit Minimum Capacity, Heat Rate at Minimum, Bid factors, Start Up Costs, Start Up Fuel, and Operating Fuel choice decisions are fully represented in this algorithm. If estimated profit over the minimum up period exceeds the economic hurdle rate for commitment (specified through the unit’s Non-cycling factor), a decision is made to commit the unit (unless the forecasted value of operation in the first hour is negative, excluding all start-up costs).

A similar process is used to evaluate the economics of shut down decisions for any non-cycling units that are eligible for de-commitment in the hour. Aurora will decide to either continue operating the unit for an additional hour, or to shut the unit down (de-commit), depending on the expected consequences (profitability) of continued operation. Those consequences are estimated by examining hours successively farther into the future, one hour at a time, until the accumulated forecasted operating results satisfy one of two alternative conditions; either accumulated value (revenues – variable costs) is a loss that exceeds start-up cost, or accumulated value is positive. Here is the process: Beginning in the first hour a running commitment resource is eligible for de-commitment, Aurora looks forward as many hours as necessary, accumulating the projected operating value of the unit, until the accumulated value (adjusted by the same unit’s non-cycling factor) is either a loss of greater magnitude than the start-up cost of the unit, or is positive. In the former case, the resource will immediately de-commit, and in the latter it will continue operation. In each subsequent hour, the same commitment evaluation is repeated. If this evaluation proceeds a full week (168 hours) into the future without resolution, then the unit will de-commit. In the case where no startup costs are specified for a unit, a comparison is made between the unit’s incremental cost adjusted downward for the non-cycling factor, current prices, and the average price forecast over successively longer horizons. At each successive horizon, if both the current zone price and the average forecast price are below the adjusted incremental cost, a de-commit decision is made; but if the unit achieves an accumulated positive value over the horizon, then Aurora will stop examining further horizons and the unit will continue operation for an additional hour.

Additionally, Aurora performs an evaluation of de-commit decisions that imposes constraints on the number of units that can shut down in any hour for a given fuel type within a given zone. The logic is applied to the set of commitment units within a zone using an individual fuel type and there is a threshold for both the number of units and the cumulative capacity that must be met before the logic is applied. The thresholds are set at a minimum of 4 commitment units with a minimum combined capacity of 2000 MW for a given zone/fuel type. If a unit set qualifies, no more than 25% of the capacity within that group is allowed to de-commit in any hour. Therefore, there is a potential hazard of using multiple fuel types to segregate resources that actually have identical fuel characteristics.

See the Aurora Self Commitment Flow Diagram for visual representation of this logic.

Decision Method 2 – Pool-Based Commitment

This option allows the use of an alternative method for unit-commitment within Aurora. This option is selected in the Operating Pools table by setting the Pool Commit column to TRUE. When this option is set to TRUE, the standard price-based self-commitment logic described above will not be used for any units within the pool. Instead, the model will use a combined reliability and cost-based algorithm to coordinate the unit-commitment for all units in a pool. This algorithm reflects the same parameters and constraints as the standard algorithm including minimum up time, minimum down time, minimum segment size, heat rate at minimum, and Startup Costs. However, the objective of the algorithm is to find a coordinated commitment schedule across all pool units which will satisfy reliability and spinning reserve requirements at the lowest estimated system cost. This method does not take into account making individual plants economically whole and therefore it is possible for units will lose money when they run since the objective is total cost minimization for the pool.

The algorithm is run prior to the first hour of each dispatch day, with decisions made for the next 24 hour period. A 48 hour data window is used to handle boundary effects. The commitment pattern from this schedule is then locked in for the next 24 hour period and used to update the hourly unit status as the simulation proceeds. This commitment algorithm can be more computationally intensive than the standard methodology and may have a minor effect on model run time. The pool level time series variables for Firm Imports and Firm Exports can be used for additional control over the unit-commitment patterns. It is also possible to specify pool membership for units which may physically exist outside of the zones included in the pool definition by specifying a Reliability Area and Reliability Share, but due to contractual obligation or majority ownership, will have commitment behavior coordinated with an operating pool.

Non-Commitment Penalties

Non-cycling units that are eligible to run in hour, but which have not had an explicit decision to commit for the hour made through the logic options described above, will generally still be made available to the system dispatch algorithm, but at a dispatch cost which includes a non-commitment penalty. This penalty will be calculated in one of two ways depending on logic options selected by the user.

A Fixed Non-commitment Penalty will be calculated as a user-specified ratio of unit full output cost. This value is added to the incremental cost of the unit to determine the dispatch cost. Note that this method will always be used for pool commitment units.
An Economic Non-commitment Penalty will be calculated as the product of the average commit hurdle shortfall over the minimum up period and one half the minimum up period. This effectively says that if the zone clearing price observed in the hour is high enough to recover at least half the economic shortfall originally expected over the minimum up period, the unit will start in the hour.

Initial Commitment

At the beginning of a study, Aurora will run the commitment algorithm, dispatch solution, and internal price forecast algorithm, in an iterative fashion to establish an initial commitment pattern for all non-cycling units. The objective is to establish internal consistency between the initial commitment status for non-cycling units, initial zone clearing prices, and the hourly internal forecast before moving beyond the first hour of the simulation. When this initial commitment process has converged it provides information about which units on the system should be committed and which should not for the first hour of the study. Because a history for unit elapsed time up and time down is needed going forward in the simulation, the model distributes the decision points for the implied startup and shutdown decisions backward in time to provide some diversity in the timing for future commit/decommit eligibility events.

Simulation Logic

Traditional Commitment Logic