|
Q1: Several RTO/ISOs have already implemented Marginal Losses into their market designs. What are the differences between the PJM implementation and the features of the other RTO/ISOs implementation?
|
A1: Below is a table providing an overview of the major features of the MISO, ISO-NE, and NYISO marginal loss implementation compared to the PJM implementation.
| Approach |
PJM |
MISO |
NY |
NEISO |
| Loss charged on average loss or on marginal loss |
Currently Average |
Marginal |
Marginal |
Marginal |
| Marginal Loss Reference |
Distributed Load Reference |
Distributed Load Reference |
Fixed Single Station Reference |
Distributed Load Reference |
| Penalty Factor Calculation |
Determined from the incremental losses calculated in an Optimal Power Flow |
Determined from the incremental losses calculated in an Optimal Power Flow |
Determined from the incremental losses calculated in an Optimal Power Flow |
Determined from the incremental losses calculated in an Optimal Power Flow |
| Allocation Method |
Allocated to Transmission Users based on load plus exports ratio share. |
Load Ratio Share within loss pools. Allocation to loss pools determined according to “loss charge” incurred from generation to load within each pool |
Over collection reduces amount of Minimum Generation and Startup Residual Uplift allocated through Schedule 1. Essentially reduces load ratio share charges |
Allocated according to Real-Time Adjusted Load Obligation |
|
| Back to Top |
|
Q2: Are penalty factors applied to Generation and Incremental offers?
|
|
A2: Yes. The penalty factors are applied to each and every location within the PJM network model. Every generator, load, inc, dec, fixed demand, price sensitive demand, etc will have the marginal loss effect associated with their location included in the clearing process as well as the resulting prices.
|
| Back to Top |
|
Q3: Will the Losses be calculated in order to carry out economic dispatch or will it be calculated post-dispatch?
|
|
A3: PJM will apply Marginal Loss in the process of commitment, dispatching and pricing. Real-time penalty factors are calculated based on the real-time SE solution. And Day-Ahead penalty factors are based on the Day-Ahead model.
|
| Back to Top |
|
Q4: Will virtual transactions have an impact on marginal loss calculations?
|
|
A4: Virtual transactions are considered in the Day Ahead clearing just as any other generator or load. Therefore, their effect on losses will be taken into consideration and they will be cleared in the same manner as generation or demand. In other words, they will be treated on an equal basis as actual generation or load within the Day Ahead market. Since virtuals do not appear in Real Time, they have no impact on Real-Time losses.
|
| Back to Top |
|
Q5: If incremental losses are a function of power injections, how are penalty factors determined for load?
|
|
A5: The penalty factors are calculated for every location within the PJM network model by determining the change in system losses for a given increase in power injection. This calculation is done for every location regardless if there is a physical generator or load at the location. Therefore, a penalty factor for every location on the system is obtained. From this comprehensive set of penalty factors, the marginal loss effect at every location, including load and generation, can be determined. Day-Ahead penalty factors are based on the Day-Ahead model and the Real-Time penalty factors are based on real-time system conditions.
|
| Back to Top |
|
Q6: How will the reference bus be chosen?
|
|
A6: The reference bus will be chosen by using a load weighted average of all load buses within the PJM footprint. This load weighted average is called a distributed load reference or “center of load” reference.
|
| Back to Top |
|
Q7: Why is PJM using a distributed reference bus?
|
|
A7: The reason for using a distributed load reference with the ability to change every execution based on actual system conditions is that this method minimizes the error introduced by the linear approximation of the actual differential equations used to model transmission loss impacts. This method is being utilized in other RTOs including Midwest ISO and ISO-NE.
|
| Back to Top |
|
Q8: How often will the reference bus change?
|
|
A8: PJM will use a distributed load reference in the calculation of Marginal Losses. Because the demand at each load bus can change with every execution, the distributed load reference is recalculated each and every execution. For the Real-time LMP calculation, this occurs every five minutes and will use the actual load MWs at each load bus for the given period the calculation is executing. For the Day-Ahead LMP calculation, this occurs every hour and will use the fixed load MW, cleared price-sensitive load MW, and a share of total losses for each load bus.PJM will use a distributed load reference in the calculation of Marginal Losses. Because the demand at each load bus can change with every execution, the distributed load reference is recalculated each and every execution. For the Real-time LMP calculation, this occurs every five minutes and will use the actual load MWs at each load bus for the given period the calculation is executing. For the Day-Ahead LMP calculation, this occurs every hour and will use the fixed load MW at each load bus.
|
| Back to Top |
|
Q9: Can a generator lose money by getting a LMP less than their offer?
|
|
A9: The PJM implementation of Marginal Losses includes the incremental loss effect of each location within the optimization. Therefore, the solution to the dispatch and the pricing will be consistent and each will consider the effect of Marginal Losses on a generating unit. As long as the unit is following PJM dispatch and determined to be eligible to set price, the dispatch instructions and pricing will be consistent. Any discrepancies between the dispatch and resulting price for units following PJM dispatch will not be a result of Marginal Losses.
|
| Back to Top |
|
Q10: Where will the distributed reference or “load center” be located within PJM?
|
|
A10: Because the distributed load reference or “load center” is a load MW weighted value, the location of this reference will change as load changes. The actual load MWs are used for the Real-Time LMP calculation and the fixed load MW are used for the Day-Ahead LMP calculation to calculate the distributed load reference. Therefore, the location can vary each and every execution. Because it is load weighted, the location of the distributed load reference will be near the load center and is at times referred to as the “load center”.
|
| Back to Top |
|
Q11: Can Marginal Losses be negative with respect to the reference bus?
|
|
A11: Yes. The marginal loss component of LMP can and will be negative for some buses with respect to the reference. If an increase in generation at the bus results in an increase in losses, then the marginal loss component of that bus will be negative.
|
| Back to Top |
|
Q12: How will the approximately $100 million of total production cost savings be distributed among the various zones?
|
|
A12: There will not be a $100 million “pot of dollars” to be allocated annually following implementation of Marginal Losses. The approximately $100 million of potential production cost savings from implementing Marginal Losses is a result of considering a generators effect on system losses. Because the dispatch will change to minimize the losses incurred on the transmission system to the extent it is economic to do so, the total MWs produced in order to serve actual load plus losses will be reduced, thereby resulting in the reduced production cost. The resulting dispatch will be more efficient and the total production cost will be reduced.
|
| Back to Top |
|
Q13: Why is there an over collection with the implementation of Marginal Losses?
|
|
A13: The implementation of Marginal Losses will include the loss penalty factor into the dispatch and pricing calculations. By including the penalty factor, losses will be paid on the margin instead of on the average. By paying for losses on the margin instead of on the average, an over collection occurs. For example, if there is energy flowing on a line because of only one party, the losses on the line may be 0.5%. If energy from another party is added to the line the losses may go up to 0.8%. If even more energy from another party is added to the line the losses may rise to 1.3%. If this process continues until all the energy from the various parties is flowing on the line, the losses may be at 4.5%. Because everyone will pay losses on the margin, they are all paying based on the 4.5% of losses instead of the average losses. Since the average losses will be less than the 4.5% Marginal Losses, an over collection occurs.
|
| Back to Top |
|
Q14: Why can’t Marginal Losses be implemented without an over collection?
|
|
A14: The implementation of Marginal Losses will include the loss penalty factor into the dispatch and pricing calculations. By including the penalty factor, losses will be paid on the margin instead of on the average. It is impossible to include the effect of Marginal Losses and not create an over collection.
|
| Back to Top |
|
Q15: Does the over collection mean that some parties over pay for losses while some under pay?
|
|
A15: No. The over collection occurs because of paying losses on the margin instead of on the average. Using the example in FAQ #12, it is impossible to determine which party was “on the line first” and would therefore get the 0.5% charge. Everyone is charged equally based on the Marginal Losses and an over collection occurs. Therefore, it is not possible to determine if a certain party over pays or under pays for losses.
|
| Back to Top |
|
Q16: If the loss component is negative does the unit look more attractive to dispatch or less attractive?
|
|
A16: If the loss component is negative, an increase in generation results in an increase of losses on the system. Therefore, the unit will look less attractive to dispatch. If the loss component is positive, an increase in generation results in a decrease of system losses and the unit will look more attractive to dispatch.
|
| Back to Top |
|
Q17: Are the loss components the penalty factors or adders to the total price?
|
|
A17: The penalty factor is a measure of the change in system losses for a given increase in power injection at each location. The penalty factors are used in the pricing to take the marginal loss effect into consideration when calculating the prices. The marginal loss effect contained within the penalty factor is turned into a loss component that is then used as an adder or component of the total price. The total price includes three components; 1) the marginal cost of generation, 2) the transmission congestion cost and 3) the Marginal Losses.
|
| Back to Top |
|
Q18: PJM has stated that they will use a linear approximation to model the incremental losses. How will PJM model incremental losses and what is the linear approximation?
|
A18: The model of losses used is a linearized model based on the sensitivities of overall system network losses to location-specific changes in power injection/withdrawal. Using this linearized model, we can write the total network losses as
 (image)
where
(image)
The linear approximation of losses as a function of net injections is made at a particular system state. The state of the power system includes an electrical network model, a topology, and a set of nodal injections and withdrawals.
|
| Back to Top |
|
Q19: Using the linear approximation for incremental losses, how will PJM calculate the marginal loss penalty factors?
|
A19: The calculation of the marginal loss penalty factors is accomplished utilizing the Newton-Raphson portion of the Optimal Power Flow function. The penalty factors are a function of the loss sensitivities calculated for the real-time state of the network. The loss sensitivities can be viewed as the change in MW at the reference bus due to a change in MW injection at a generator bus. Such sensitivity terms can be determined for all busses by solving the following equation using Gausian Elimination:
(image)
If a unit j is connected to bus i then unit j’s loss sensitivity is given by
(image)
and the corresponding loss penalty factor is
(image)
|
| Back to Top |
|
Q20: What is the pricing formulation that PJM plan to implement for Marginal Losses?
|
A20: Given the linear approximation of losses discussed in FAQ #21 and the calculation of the marginal loss penalty factors in FAQ 22, we can define the formulation for the calculation of LMP inclusive of losses as
(image)
|
| Back to Top |
|
Q21: Does PJM use a DC transmission model in the dispatch algorithm? If yes, that would indicate the reference bus in the DC model makes up for the total system loss, which will lead to a different load flow than the state estimator. How does PJM deal with inconsistent line flows between the DC model and the state estimator?
|
|
A21: PJM utilizes an incremental dispatch algorithm that begins with the full AC powerflow solution from the state estimator. Generation is dispatched to meet load plus losses and the error that would be introduced from using the reference bus to make up system losses is therefore not introduced.
|
| Back to Top |
|
Q22: How will new assets be incorporated into the model?
|
|
A22: New assets are included in the network model and since the incremental loss factors are calculated each 5-minute execution from the network model, their effect is automatically taken into consideration.
|
| Back to Top |
|
Q23: How will verbal dispatch instructions or off-cost dispatch affect the model?
|
|
A23: The UDS solution will consider marginal loss effects for economic dispatch, regardless of whether there are binding transmission constraints in the solution resulting in off-cost dispatch. Because consideration of marginal loss impacts will better represent the physical system and improve dispatch efficiencies, verbal dispatch instructions should be reduced.
|
| Back to Top |
|
Q24: Will Marginal Losses be applied to both Day-Ahead and Real-Time dispatch? How will this affect the allocation methodology? What happens when RT power flows are significantly different from a modeled Day-Ahead flow?
|
|
A24: Yes – Marginal Losses will be applied to both Day-Ahead and Real-Time dispatch and pricing. The effect is described in the formulations given in FAQ 19. The Day-Ahead and Real- Time calculations will look at the marginal loss effect independent of the other. Since balancing settlement is based upon deviations from Day-Ahead scheduled quantities, Real Time-prices account for the difference between Real-Time and Day-Ahead flows.
|
| Back to Top |
|
Q25: When will PJM begin posting of the LMP information with Marginal Losses included?
|
|
A25: Day-Ahead Market Trials where PJM will post Day-Ahead Market results with Marginal Losses included are scheduled to begin in late March or early April.
|
| Back to Top |
|
Q26: Will PJM publish Penalty Factors for all units in the RTO for a Reference Case?
|
|
A26: No, penalty factors can be calculated based on LMP prices that are posted.
|
| Back to Top |