We now know what balancing energy is, how it is provided, who provides it, who controls it, who uses it and what other basic system design elements influence its overall usage. There’s nothing else left than taking a deep dive down the rabbit hole of the methodology of its calculation in our selected three countries. Similarly to Alice, we shall be rewarded with the opportunity to explore a wonderland where it is really easy to lose track of time and space.
Balancing energy is the delta (or the difference) of plan and actual consumption or production. Therefore, the definition of positive and negative balancing energy will depend on which goes first in the equation for generators and consumers: the plan or the “fact”, i.e. do we subtract the fact from the plan, or the plan from the fact, when we are calculating the delta? Because I’ve been told a picture is worth a thousand words, here’s a graph that illustrates what method is used in the German and Hungarian system.
The blue line represents the schedule (plan) of a consumer in MW for a day. The purple line is the actual consumption in 15 minutes increments (remember the imbalance settlement periods?). Firstly, it is visible that the schedule resembles a stairway with steps, while consumption is a continuous line. This is because the costumer in this example is not able to buy products to cover (or hedge) its load with 15 minutes granularity. Instead it is trying to estimate it with so called “hourly products”. Because of this difference, inevitably, balancing energy is needed. The orange bars show the amount of balancing energy needed in periods where the consumer under planned its usage, while the green ones indicate over planning. If the delta is fact minus plan, the orange area is a positive number, because the actual consumption was higher than the plan. Similarly, the green area will be a negative number, because actual consumption was lower than the plan. For generators, the equation will be the other way around: higher output than scheduled will be a negative number; lower output will lead to a positive number. The same logic applies to individual market players as to the whole system. There is too much generation or too little consumption in a system with a negative balance (long system), while a positive balance refers to insufficient generation or over-consumption (short system). The Australian approach is exactly the opposite: a long system (or market player) will be assigned with a positive number, a short system (or a market player) will receive a negative number. This is because the order of variables in the equation describing the deviation from plan is switched around.
Armed with all this knowledge we can finally proceed to look at actual the data and make sense of it. As I previously hinted, gate closure times, lead times, the length of imbalance settlement periods and liquid intraday markets (where appropriate) play a major (but not an exclusive!) role in the volume of balancing energy used in a given system. Here’s a refresher on the numbers, plus some new info:
|day-ahead gate closure||14:30||14:30||12:30|
|intraday gate closure||60 min||15 min||1-2 min|
|lead time||60 min||15 min||n/a|
|imbalance settlement||15 min||15 min||4 seconds/5 minutes/30 minutes|
|average demand (GW)||4.8||54.8||21.7 (8 for New South Wales alone)|
|total yearly consumption (GWh)||41,885||477,490||190,968 (70,079 for New South Wales alone)|
|total balancing energy used (2016, GWh)||666||3,232||unknown1)Australian data on balancing energy volumes is rather hard to use and to interpret, partially because of the sheer volume of it. After all, for imbalances the granularity is 4 seconds, even if it is averaged over a five minute period. As a result, Australian data is not further analysed in this post.|
|Yearly aggregated accuracy (%)||1.6||0.7||N/A|
Because markets differ very much in size, an absolute value for balancing energy used would not be particularly useful. Although we may get a hint of what’s coming just by looking at how average demand compares with total balancing energy used in a year. It is, however, possible to create a comparison between these jurisdictions. For that, let’s simply divide the volume of system imbalance, or used up balancing energy with the total system demand in each imbalance settlement period. This should give us a comparable percentage. From now on, I refer to this percentage calculated on a 15 minute basis as “accuracy”. The chart below shows that for the Hungarian system.
One more thing. If we are looking for accuracy, we should also introduce a notion called the mean absolute percentage error (MAPE). Using MAPE is necessary, because as seen in the 15 minute graph above, positive and negative values both occur in our data. Positive(orange line) in our case refer to a short system where positive balancing energy is needed, while negative (blue line) refers to a long system, where negative balancing energy is needed. For instance positive 5% means that there was a shortage in the system, and the amount of balancing energy utilised in that 15 minute interval was 5% of the total demand in the same interval. If we did not use MAPE, negative and positive percentages could cancel each other out to a certain extent, therefore, we’d be led to believe that the forecast was more accurate. However, a simple, arithmetic average is also useful for other things. I’ll explain that in a bit. The Hungarian system had a median MAPE of 1.29% and an average MAPE of 1.59%. Average shortage/surplus was 0.114%. The biggest outliers are -19% and +14% on one occasion. This does not seem too bad … until we take a look at the Germans.
The greatest surplus in their system was only -6% and the shortage +5%. Median MAPE 0.54%, average MAPE is 0.68%. Average shortage/surplus was 0.334%. Notice that Germany outperforms Hungary in every category except for the average shortage/surplus, where it does almost three times better. What does that mean? Well, the other indicators derived from MAPE show accuracy, while average is better at showing whether the system more skewed toward shortages. And that is exactly what we see here. Although in Germany there is relatively less balancing energy used than in Hungary, what do gets used tends to be positive rather than negative.
If we accept median MAPE as a good measure of accuracy in planning, the German system does a 2.4 times better job than the Hungarian one. Although it does not explain everything, it is hard not to notice, that Germany has a 4 times shorter gate closure than Hungary. Germany also has something that Hungary does not: a liquid organised intraday market. HUPX opened its intraday market segment in March 2016. But it did not, to date, gained enough liquidity to offer sufficient means for market players wishing to correct their schedules at the last minute. I’ll address the issue of liquidity in the next and final blog post. I know it is a far stretch, but it is interesting nevertheless to see what would happen to those nice and tight MAPE indicators of Germany if there was no liquid intraday market present. It is possible to approximate and simulate the “what-if” by adding the quarter-hourly traded volumes on the German intraday to the used balancing energy volumes. The median MAPE is suddenly becomes 1.32% (slightly higher than the Hungarian!) and the average MAPE is 1.45%. Both numbers are very much similar to the Hungarian ones. What a difference an ID2)ID is short for intraday, and exchange traded intraday markets are commonly referred by this acronym. makes!
Lábjegyzetek [ + ]
|1.||↑||Australian data on balancing energy volumes is rather hard to use and to interpret, partially because of the sheer volume of it. After all, for imbalances the granularity is 4 seconds, even if it is averaged over a five minute period. As a result, Australian data is not further analysed in this post.|
|2.||↑||ID is short for intraday, and exchange traded intraday markets are commonly referred by this acronym.|