Rules for Balancing Generation and Load
The control mechanism for the portable generator in chapter 1 was a governor that maintained a constant speed. This mechanism may work for small systems, but it does not work for the grid. No single generator can control the frequency since the system is so large (the peak load of the eastern grid in 2007 was 638,500 megawatts). So the control mechanism for matching generation to load is not one that relies on frequency control at all. Matching generation to load is done in a distributed control scheme, wherein defined areas (subsets of the overall grid) must match their generation to their own load. The idea is that if each area is in balance, then the whole grid will be in balance. Or, said another way, if some areas are generating less than their load, then others will be generating more than their load—so the grand total of the entire grid will be pretty close to balanced and the frequency will be close to 60 Hz.
The grid is divided into these areas by definition. That is, it is not actually “divided,” but by agreement with neighbors entities that control the grid define the equipment that will be contained in each area. Prior to 2005, these areas were called control areas, but now they are called balancing areas, which is a bit more descriptive. The rules state that all grid equipment, load, and generators must be within a defined balancing area that is controlled by a balancing authority (BA), so balancing the entire grid is controlled by a number of BAs.
These balancing areas typically fall where ownership of the grid falls, although there are many exceptions. They are also closely aligned with geographic areas, such as the certificated areas that the states have defined for IOUs, although again there are numerous exceptions. At my last count, there were 105 BAs in the eastern grid. Within a BA there can be multiple generation companies, transmission owners, and distribution companies. (However, Texas has only one BA in its grid.)
Each of these BAs is connected to other BAs by transmission lines that cross the BA boundary and then connect to other BAs. In fact, most of them have multiple connecting lines since the grid is so highly interconnected. Electric current freely flows on these lines, as the law of physics dictates, so think about a large geographic area with lines connecting to other areas, such as in diagram 2-1. To simplify the example, I show only one line connecting BA-1 to each of the three other BAs. Each of these lines has metering that can give a real-time signal of the current and voltage, and therefore the power flow is a known quantity. (As I go through this example, keep in mind that there is nothing other than the law of physics controlling the flow in each line. Also keep in mind that the example is only showing power flows between BAs. Internal to each BA are power flows of generation to load that are probably much higher than the flows on the tie-lines.)
In my example, line 1 has 100 megawatts (Mw) flowing out of BA-1; line 2 has 50 Mw flowing out of BA-1; and line 3 has 150 Mw flowing into BA-1. So if we simply total all the outgoing flows with all the incoming flows, we can determine if the BA is balanced. In this case it is in balance, since all the flows in and flows out total to zero. What is interesting here is that the system frequency doesn’t matter; also, the total load/generation in the BA doesn’t matter. All that matters is the sum of the flows, which in this case is zero, showing that there is no balance error.
Diagram 2-1: Balancing Generation and Load
I will take this example one step further and introduce what happens if the generation and load are not balanced in the area. Assume the flows in diagram 2-1 are 100 Mw out on line 1, 50 Mw out on line 2, and 160 Mw in on line 3. There is a balance error of 10 Mw. Since the power is flowing into BA-1, the generation in the area is short by the 10 Mw error. To remedy the error, the BA-1 controller sends a signal to certain generators equipped with control capability (called regulation control), signaling to those generators to increase their output slightly in order to make up the 10 Mw difference. Those generators increase output slightly, and then the balancing error disappears.
This process is continued day and night at all BAs. These controls are done automatically, with a balancing error calculated every few seconds and with signals going out to generators as needed. Keep in mind for this example that the frequency of the grid and whatever else is happening in other balancing areas doesn’t affect what this particular balancing area is doing. Even the total load in BA-1 is not a factor. (A little later I’ll bring up an important exception to that concept.)
I would like to go a little deeper in order to explain how wholesale purchases and sales of electricity are implemented on the grid. Starting with the example above, also include the datum that one of the utilities within BA-1 in diagram 2-1 has purchased energy from a utility in BA-4. Just to make the numbers work out, the utility in BA-1 has purchased 10 Mw of power for one hour (10 Mwh of energy). How is that transaction going to actually happen on the grid?
The process to implement a wholesale transaction is really quite simple. What happens is that the balancing authority in area 1 inserts a “schedule” of 10 Mw received for the hour into their computer calculation, and BA-4 inserts a schedule to deliver 10 Mw for the hour. This means that BA-1 expects to receive the 10 Mw, so the flows on the lines in the example above are as follows: 100 Mw out, 50 Mw out, and 160 Mw in, indicating that BA-1 is now in balance. That is, BA-1 expects to receive 10 Mw from the grid, and they do receive it, since the sum of all the flows equals a 10 Mw receipt. (Keep in mind that the 10 Mw schedule from BA-4 to BA-1 would not necessarily flow on line 3 as in this oversimplified example. It would more likely flow on all the tie-lines into BA-1.) Transactions like this are constantly being implemented by almost all the BAs on the grid.
I need to introduce an important control factor called frequency bias, which helps to maintain the reliability of the grid. Remember that all the BAs are doing what they can to maintain generation and load balance in their areas, although there is no overall controller. System frequency is fluctuating around 60 Hz, although no one is trying to control the frequency. When a large generator trips (which, as I’ve said before, happens almost daily on the eastern grid), the frequency on the entire grid decreases. All the BAs in the grid (with the exception of the BA that lost the generator) will experience an increased outflow because their loads will decrease with the lower frequency—so they will all be overgenerating. With the controls I have described so far (that is, controlling generation to reach a net zero on a BA’s tie-lines), there is a concern that the frequency could continue to decay, since this outflow from each BA would be curtailed. So there is a need to add a factor that allows for this flow.
In addition to the increased outflow due to the reduction in load, there is an increase in generator output due to the lower frequency. In chapter 1, I mentioned that the governor in a portable generator could be set to maintain a target speed. It turns out that most of the grid’s large generators have that capability. So as these generators sense the frequency decline, their governors can (automatically) boost the output of the machine, thereby helping to arrest the frequency decline. This governor boost reaction to a low frequency happens with most generators on the grid, and it happens within the first minute after the loss of generation anywhere within the grid.
Backing down generation throughout the grid is not desirable when frequency is low, as generator boost during low grid frequency is a good thing, so therefore the frequency bias is included in the programming for all BAs. Frequency bias allows for some additional power output on a BA’s tie-lines proportional to the grid frequency. The frequency-bias factor doesn’t provide for the grid frequency to recover to the preferred frequency, but it does provide assurance that the decay in frequency will not become worse. To put into perspective what this factor amounts to in power output, I will take a look at the standard. The standard requires that the bias setting should be no less than 1 percent of the system load for a change of 0.1 Hz. For a system of 20,000 Mw, the bias would be set at no less than 200 Mw, which, when spread over a large number of generators, is not that much extra generation. On the eastern grid, a change of 0.1 Hz is huge; this much frequency change is very unusual.
What about the BA that just lost the generator? The standards require the deficient BA to make up the loss in generation, in any way possible, within fifteen minutes of the unit trip. There are several options for doing this, and some of them are better than others. The fifteen-minute requirement pretty much rules out starting any new generation, unless the BA has some small gas turbines available that can take significant amounts of load in this short time frame. The larger gas turbines that have been built in the last ten years are a lot more efficient than the older, smaller turbines from the 1960s and 1970s, but they take too long to get up to speed and also take load to help out in this situation. Any generator within the BA that is online and available to produce more power can be used, but often there is not enough unused capacity available in the units online to replace a large unit. Hydro units are excellent for this purpose if they are available and not already running, since they can get up to full load in a few minutes after being started. Also, the BA can have an arrangement to buy or even borrow the needed energy for some period of time from another area until the BA can start another large generator. Typically, this kind of arrangement (called reserve sharing) is done ahead of time so that it can be quickly implemented at the time it is needed. But the important point here is that somewhere in the grid, generation must be increased to make up for the loss of generation that is the result of the unit trip.
This discussion about balance error and BAs is very important to understanding how the grid is controlled. To summarize: The balance error is calculated based only on a net sum of all the flows on all the tie-lines for a single BA, adjusting for all wholesale transactions across the BA border, which are entered into the computer as schedules and are then adjusted slightly again based on grid frequency. This calculation yields a number that may be positive or negative. The control of generators within each BA is performed based on that calculated factor. We call this balance error value the area control error, or ACE. If it is positive, then it indicates that the BA is generating more than necessary. In such a case, some units will be directed (automatically, of course) to reduce their output. And if it is negative, then the BA is generating less than necessary and some units will be directed to increase output. The standard established by NERC gives guidance for measuring how well each BA matches its generation to its load. The algorithm is complicated, but the point is that some balancing error in real time is expected and allowed. It would be virtually impossible to maintain an exact balance all the time.
If anyone looked at a grid frequency trend recorder for the last couple of hours on most days, that person would see a wavy line that goes over 60 Hz and under 60 Hz continually. When the grid is at 60 Hz, we say that the generation and load are balanced, which really means that for an instant of time the load is met exactly by the generation and the frequency happens to be 60 Hz. If the load and generation are said to be imbalanced, what we really mean is that the frequency has deviated from 60 Hz. If the frequency is higher than 60 Hz, then the generation output at that instant is too high. Conversely, if the frequency is lower than 60 Hz, the generation output is too low. Fortunately for us all, electrical equipment (generators and load) is designed to handle some deviation in frequency. More important, the inherent characteristic of most of our load devices is that the energy required at a lower frequency is slightly less than the energy required at a frequency of 60 Hz. The reason I say this is fortunate is that the small imbalance between generation and load is simply absorbed by the ability of the load to increase or decrease with the frequency. Remember that all the electricity generated is used instantly. This allows the system to be in equilibrium and yet stable, which it is, generally.
It is interesting to note that this system of multiple BAs controlling to their own balancing equation makes it very difficult to identify in real time which area is causing a problem. Say for example that the frequency on the grid has been low all day. Each BA in the grid has been supplying the frequency-bias energy as required, and by now they all want to know who is getting this energy and when they are going to get the low frequency straightened out. We’ve seen this happen many times. There are at least two obvious ways that this can come about, and either one is tough to identify.
First, say there is a BA with a metering error on a tie-line with a neighboring BA. Actual flow in this example is 100 Mw, whereas the meter shows 10 Mw. The BA uses that meter, along with all the other tie-line meters, to calculate a total flow in or out of the area, so whatever it is that they expect to see is off by 90 Mw. In effect, the BA is receiving 90 Mw more than the computer realizes. This goes on until someone catches the error. This meter error would result in the grid frequency being a bit slow, since there is an undergeneration in the grid of 90 Mw.
Second is a case where one BA thinks there is a transaction from a generator located in another BA of 100 Mw, but the other BA thinks it is 10 Mw. The outcome is the same as in the first example. This imbalance could simply be the result of an entry error somewhere, or, as we saw in the late 1990s, it could be the result of a dispute wherein the two parties simply disagreed and left it at that. I know it sounds ridiculous, but this actually happened for several months, as I recall. During that time, every other BA in the eastern grid was supplying the makeup energy to boost frequency and was not being paid for the energy. The energy supplied over several months turned out to be thousands of megawatt-hours. What several BAs finally did was simply to remove their frequency bias until the matter was resolved. By removing their frequency bias, they were no longer supplying energy to the grid for which they were not being paid, but at the same time they were putting the grid at some additional risk.
I have tried to provide a basic description of how generation and load are balanced on the grid in real time. This discussion can be expanded to include the processes that take place to plan the present day, the next day, and the upcoming week. Without going into detail, I’ll say that each BA would be responsible for developing viable operating plans for the time horizon of one week in the future. This plan would include load forecasts, generator availability plans, outages, purchases, sales, and any other inputs that may need to be on a BA’s radar.
The rest of the discussion on how the grid is controlled will focus on methods of preparing for and dealing with unplanned events, and dealing with the competitive market.