Regional transmission organizations (RTOs) or independent system operators (ISOs) dominate the major power grids of North America, with the notable exceptions of the Southeast and Pacific Northwest. The Federal Energy Regulatory Commission (FERC) has been a strong advocate of RTOs as a way of opening access to the power grids and improving system reliability. The power grid obviously requires active, hands-on management to ensure system reliability, but even that has not been sufficient to prevent major outages. New rules put in place following the last major blackout put more teeth into reliability.
The purpose of this article is not to criticize system reliability but to highlight the more pervasive challenge today and for the future: Controlling the cost impact of decisions by grid operators on energy market participants. Here, as elsewhere, the law of unintended consequences is alive and well, as procedures put in place to avoid outages at almost any cost have the consequence of raising the overall cost of operations through rapidly rising congestion costs.
Consider a few examples.
Economic dispatch generally means choosing the lowest operating cost electricity generation supplies to meet electricity demand. For example, assume a 1,000-MW electricity demand in an hour, with five generating units to choose from. Units 1, 2, 3, and 4 are each 250 MW and cost $20/MWh to operate. Unit 5 also is 250 MW, but costs $30/MWh to operate. Using an economic dispatch approach to resource allocation would give preference to the least-cost unit sufficient to satisfy the requirement. Thus, in this example, economic dispatch would result in the decision to operate units 1-4 to meet the 1,000-MW load, leaving the more expensive unit 5 offline.
To make this economic selection of resources, grid operators need to have a transmission system capable of delivering the power from units 1 through 4 to the load. Continuing with our previous example, assume that the generation and load are connected via transmission according to the following:
With this arrangement, and all lines in service, the system can allow each of generators 1, 2, 3, and 4 to run full out to meet the 1,000-MW load. An operating reserve requirement equal to the largest single generator contingency (i.e., the possible low of one of the 250-MW generators) is covered with the 250-MW spinning reserve unit. Generator 5, which is more expensive to operate, is not needed and can be shut down. This dispatch pattern can be considered to be a security constrained economic dispatch (SCED), since the system is designed to survive the loss of the largest single generator through the provision of spinning reserves with the 250-MW spin generator.
System operators also consider whether the loss of a transmission line might cause a problem. If the loss of any of the operating generators 1 through 4 would cause the system to become unstable, causing cascading blackouts (within less than a minute of the outage), something must be done in advance of the outage to ensure the system would not become unstable should the outage occur. For this example, we will assume that such an unstable condition would not result from any outage contingency.
Even without a system-stability concern, system operators still consider whether the loss of a transmission line might cause a problem (e.g., a thermal overload of remaining transmission lines). As can be seen from this example, with both generators 1 and 2 operating at full capacity to serve the load, there appears to be adequate transmission to move the 500 MW of power from these two units to the load, since 800 MW of thermal capacity is on the cut path between the two busses. Of course, the system operator will want to run a load flow to assure that no single line gets overloaded. Assume we have run that load flow and verified that none of the 5 transmission lines overload.
Next, assume we want to check the loading on the lines if the 280-MW line 1 trips off. As can be seen, even without line 1, there still is 520 MW of thermal capacity on the cutplane between the two busses. However, if we run the load flow, we might discover that the loading on the lines does not distribute according to the capacity of the lines. We might find that the 20-MW line 5 in fact loads to 25 MW. This exceeds its thermal limit.
Is this a problem that would prohibit us from performing the economic dispatch of running generators 1 through 4 to meet the load on this hour? Recall that while we are studying the possibility that the 280-MW line 1 trips off, we have no indication that such a trip will occur. In fact, historic data suggests that in all probability, the 280-MW line 1will not trip.
The choices available to the grid operator are the following:
1) Because of the possible tripping of the 280-MW line 1 and resulting overload on line 5, redispatch the system to back down generator 1 or 2 and run the more expensive generator unit 5;
2) Allow the economic dispatch of running generators 1 through 4, but open line 5 in the middle so that it becomes two radial lines. This way, if the 280-MW line 1 trips, then line 5 will not overload (assuming we have tested that lines 2, 3, and 4 can handle the 500 MW of transfer if line 1 trips);
3) Allow the economic dispatch of generators 1 through 4, but “arm” generator dropping on generator 2 so that if line 1 trips, relays will disconnect generator 2. We will need to test to be assured that the 250-MW spinning reserve unit will pick up to cover the loss of generator 2 without any resulting instability. Then the more expensive generator 5 could be called into operation;
4) Allow the economic dispatch of generators 1-4, but have operators take action to reduce generation on generator 2 if line 1 trips. We know that the loss of line 1 results in thermal overloads on one or more remaining lines. We also know that thermal overloads on the line mean that the line tends to warm up and “sag” as a result of thermal overloads. We also know that this phenomenon of warming up and “sagging” takes some time (e.g., 15 minutes before the line sags below acceptable clearance levels). Therefore, if line 1 trips, operators have a little time before the loss of line 1 manifests itself into a problem on the grid. Therefore, well-trained system operators have time (once they get the alarm that line 1 has tripped), to take action to avoid the problem on the grid. For example, the system operator could manually cause generator 2 to reduce its generation, with the make-up power coming from the spinning-reserve unit and then call on the more expensive generator 5 to allow recovery of the spinning reserve;
5) Allow the economic dispatch of generators 1 through 4, but have operators take action to open line 5 in the middle if line 1 trips. As indicated above, if line 1 trips, operators have a little time before the loss of line 1 manifests as a problem on the grid. Therefore, well-trained system operators have time (once they get the alarm that Line 1 has tripped) to take action to avoid the problem. For example, the system operator could open line 5 in the middle; or
6) Allow the economic dispatch of generators 1 through 4, but have operators take action to: (a) reduce generation on Generator 5; and (b) simultaneously drop load to offset the loss of generation. Then immediately call on the more expensive generator 5 to start up to pick up the load that was dropped.
Among these 6 choices, only choice number 1 actually requires us to move to a less than optimal generation dispatch in anticipation of a possible (albeit very low probability) loss of a transmission line.
Are we redispatching unnecessarily in the name of security constrained economic dispatch?
Prior to the formation of ISOs or RTOs, owners of transmission lines generally would not cause uneconomic dispatch prior to the low-probability loss of a line. Instead, these owners reasonably would be expected to choose to use one or more of steps 2 to 6 above as the plan to deal with a potentially problematic, but low-probability, line loss.
With the establishment of ISOs or RTOs, we are finding that redispatch is being done more often to deal with a potentially problematic, but low-probability, line loss. This redispatch is being done, apparently, to demonstrate that the ISO is performing security constrained economic dispatch. This is the transmission system equivalent of “cover your rear,” but it has the unintended consequence of imposing more cost on the transmission system than the simple, common-sense solution illustrated above.
The redispatch often taking place is manifesting itself as “congestion charges.” As a result, the existence of high levels of “congestion charges” is causing stakeholders to conclude that some method needs to be designed to properly assign these “congestion charges” (e.g., locational marginal pricing [LMP]-based congestion charges). In Global Energy’s view, these congestion charges can be largely avoided simply by choosing one of the steps 2-6 above rather than choosing step 1 when N-1 contingencies might indicate a problem.
Example 1: Before ISO. Four units of coal-fired generation (Colstrip Units 1 through 4) were built in the 1970s and 1980s in eastern Montana primarily to provide power in Washington, Oregon, and western Montana. Owners of the plant needed to build (or cause to be built) a new transmission system sufficient to move the power. It was determined that a design that included two 500-kV transmission lines would not be sufficient to move all the power under certain fault conditions. Rather than spending more money for a more expensive system and rather than simply not running the plants at nameplate capacity, the owners chose to install a generation-dropping scheme that would trip one of the units if the particular N-1 contingency actually occurred. This is an example of non-ISO systems avoiding redispatch cost for low probability N-1 contingencies.
Example 2: Before ISO. In the Pacific Northwest, there often is a desire to move more power from British Columbia (BC) to the United States to displace higher-cost generation in the United States with lower-cost hydro generation in British Columbia. It became apparent that either more transmission would need to be built to accommodate higher transfer of power (under possible N-1 transmission outages) or generator dropping would need to be added to generators in BC to drop generation in the event an N-1 contingency occurs. These generator-dropping procedures have been agreed to and installed by the owners of the transmission and generation. This is an example of non-ISO systems avoiding redispatch costs for low probability N-1 contingencies.
Example 3: Before ISO. Southern California power providers often like to import cheaper power from the Pacific Northwest rather than operating more expensive generation in Southern California. For example, the Northwest/ Southwest DC line that runs from the Columbia River to Southern California can carry nearly 3,000 MW of power. If Southern California can load this line with 3,000 MW of low-cost power, it can avoid running higher-cost resources in Southern California. This is a least-cost benefit solution that appears to benefit all.
However, what if the 3,000-MW DC line trips off during a peak hour? Would that cause a problem? The answer is well known to be “yes.” But the loss of this line during a peak hour is a very low-probability event. It would be very expensive to choose to run units in Southern California every day rather than importing 3,000 MW of power from the Northwest. If the low-probability event actually occurs during the peak hour, the reasonable and most cost-effective course of action almost certainly would be to drop load for a short period of time until more expensive plants in Southern California can be brought into action. This is what happened one day in late August 2005. While the dropping of load caused some inconvenience to some customers, the plan has saved huge amounts of money for California ratepayers. This load-dropping plan, which was put in place before the California ISO came into existence, still is being used by the ISO.
Example 4: After ISO. A utility in PJM tells a story about how congestion charges to the company changed after the PJM ISO decided to be more concerned about N-1 contingencies. This company was being charged about $2 million per year for congestion charges. Then one year the ISO decided it needed to start worrying about the possible overload of 69-kV lines under a low-probability, but possible, outage of a higher-voltage line. It appeared the 69-kV line would overload under such possible outage potentiality, so the ISO decided it needed to redispatch the system (even prior to the outage). This change alone appears to have raised the congestion charges to the utility from $2 million per year to about $12 million per year.
Example 5. After ISO. Approximately 755 MW of nameplate wind generation was developed in the vicinity of McCamey, Texas. Three lines existed that move power out of that area. Each has a thermal capacity of 250 MW. However, the Electric Reliability Council of Texas (acting as the ISO) determined that they needed to consider the N-1 contingency of losing one of the lines. Therefore, the wind generators were told not to generate more than 500 MW. This meant that Texas ratepayers were not getting the benefit of zero-cost wind that could have been moved because of a concern about a possible low-probability line outage.
When ERCOT was asked if it had ever considered doing something like choices 2 through 6 above, it indicated it had not because it had decided to build new transmission. However, lots of “congestion charges” were incurred prior to getting the new transmission built!
Knowing that treatment of transmission contingencies can have an enormous impact on the amount of congestion charges (redispatch charges) and LMPs, Global Energy has sought to get clarity from ISOs/RTOs about the factors to be assessed for every N-1 contingency possible before deciding what dispatch to accept. Further, we have sought clarity on what lines will be monitored. For example, will the ISO be concerned about overloading even very low-voltage but parallel lines, such as 69-kV lines? We have sought clarity on when an ISO would decide that choices 2 through 6 might be acceptable alternatives to the redispatch required under choice 1 if operating studies show that a low probability N-1 contingency would be problematic for a particular desired dispatch pattern.
What we have learned is that ISOs either don’t know, have not considered, or are not willing to provide clarity to these matters. Most said they let their operators make these decisions on a case-by-case basis from one day to the next.
Without this clarity, and without knowing what decisions operators should be expected to make in such circumstances, reasonable concerns can be raised that the system is being operated in a manner that results in more costs than incurred before the ISO’s involvement.
As ISOs report more and more congestion cost (i.e., redispatch cost), one can’t help but wonder if the increased congestion cost is being driven unnecessarily by decisions that fail to adequately consider the tradeoff between reliability and cost. It is obvious that from the ISO standpoint the safer course of action is to spare no cost to avoid an outage.
In our opinion, energy market participants should be asking the ISOs and RTOs the following questions:
1. Are we spending too much time figuring out how to allocate congestion costs (e.g., by using LMP) when we should be trying to figure out how to eliminate the congestion cost?
2. Who is checking to ensure that we are not redispatching when there are other more economic choices to be made in light of a low-probability N-1 congestion problem?
Only then will we know whether we are truly getting the least-cost, best-fit economic dispatch intended.