Counterintuitive Flows in 2023 – part 3 of 3

Over the past two weeks, I published the first two articles in a series of three, focused on understanding and analysing counterintuitive flows in the European day-ahead electricity market.

In part 1 (link) I introduced what counterintuitive flows are and why they occur.

In part 2 (link) I compared intuitive and counterintuitive commercial exchanges.

Today I will analyze market data from 2023, identifying trends and patterns by (type of) regions – where region means both formal regions (e.g. CCRs) and area by type of grid (radial or meshed).

Before delving into explanations, let’s visualize what all the abbreviations for capacity calculation regions (CCRs) and bidding zones (BZs) in this article mean.

Data Analysis

Let’s have a look at the yearly aggregated data and try to recognize some of the results we have already seen in the situations described in part 2 (link).

Note: for ease of understanding, I will use four labels to describe the market outcomes

• Price convergence: Zero price spread, with commercial exchanges
• Congested: Non-zero price spread, with commercial exchanges in the intuitive direction
• Counterintuitive: Non-zero price spread, with commercial exchanges in the counterintuitive direction
• Outage: Non-zero price spread, without any commercial exchange. Please note that this is compatible with an outage, but it does not necessarily mean that there was an actual outage, there might be more complex reasons at play.

We can distinguish the European market in regions with similar grid topology and the nature of allocation regime:

• Explicit allocation in a meshed grid – Switzerland
• Price-coupled NTC in radial regions - SWE, SEE, Baltic, GRIT CCRs
• Price-coupled meshed grid with flow-based allocation – Core CCR
• Price-coupled NTC with lineset constraints - Nordic and Hansa CCRs
• Price-coupled NTC with allocation constraint - Italy North CCR

Explicit Allocation - CH

As introduced earlier, explicit allocation almost always results in suboptimal allocation of cross-zonal transmission capacity, because usually some capacity is allocated and nominated in both directions on a bidding zone border. In the figure above we see that the Swiss borders never reached price convergence in 2023. This is likely because the explicit auctions take place earlier in the day than SDAC and the CH energy clearing price is set for CH at 11:00 CET (closing time of the Swiss auction on EPEX), while the neighbouring bidding zones take part in SDAC one hour later.

The Swiss borders were congested during 70-80% of the MTUs, while counterintuitive flows were observed 15-30% of the time. As seen in part 2 (link), counterintuitive flows with explicit allocation occur because traders “bet” on the day-ahead prices in the relevant adjacent zones, and sometimes they get it wrong. The CH-NORD border shows fewer instances of counterintuitive flows because, historically, Italy Nord has been a net importer, such that the CH-NORD border is easier to forecast. Additionally, NORD is (not yet) part of a flow-based region, where prices are strongly linked across bidding zones, making price forecasts generally difficult.

Price-coupled NTC in radial regions

Price-coupled radial regions (SWE, SEE, Baltic, GRIT) typically show intuitive flows (i.e. either with “price-convergence”, or “congested”), with the exception of LT-PL, most likely due to the Polish Allocation Constraints (see part 1 for an explanation). Of these bidding zone borders, ES-FR, GR-SUD, EE-FI, LT-SE4 and LT-PL show price convergence during 30-40% of MTUs and 45-60% MTUs with congestions, while the remaining borders show price convergence almost 95% of the time.

We also see “outages” (either planned or unplanned) on the HVDC interconnectors between GR-SUD, PL-LT, LT-SE4

The very few instances of counterintuitive flows on GR-SUD are most likely due to ramping constraints (thanks to Yves Langer for pointing this out).

Price-coupled meshed grid with flow-based allocation – Core CCR

The electricity grid in Core CCR is highly meshed. We can see that:

• Price convergence occurs approximately 30% of the time on all bidding zone borders (excluding the Polish Core bidding zone borders, where these occur ~15% of the time, again due to the Polish allocation constraint)
• Congestions occur during ~40-60% of the MTUs. The border most often congested in 2023 was DE/LU-PL, followed by BE-DE/LU, with respectively ~62% and ~60% of congested MTUs
• Counterintuitive flows occur ~10-40% of time. The least counterintuitive borders were the German borders: DE/LU – BE, AT, NL, FR with respectively ~7%, 14%, 16%, 16%. The most counterintuitive ones where PL-SK, PL-CZ, HU-HR, HU-SK with respectively ~40%, 34%, 33%, 31%. Again, the Polish allocation constraints are the likely cause for (at least) the first two of these. Note: The outcomes of flow-based capacity calculation (applied in Core CCR) are prices and net positions, while commercial flows result from calculations based on SDAC scheduled exchange methodology. Therefore, while “price-convergence” is well-defined, counterintuitive vs congestions are not. Thanks Joost Greunsven for pointing this out.
• The “outage” situation (defined as absence of scheduled commercial flows with non-zero price difference) occurs mostly on the BE-DE/LU border, since this is the only bidding zone border consisting solely of an HVDC interconnector

The bottom line is that most bidding zone borders in Core behave in a rather homogeneous way, with the exception of the Polish borders.

Price-coupled NTC with lineset constraints - Nordic and Hansa

In the Nordic and Hansa CCRs, the situation is somewhat in between Core and radial areas, including the influence of constraints (i.e. the Linesets introduced in part 1 - link). The grid in the Nordic synchronous area is fairly meshed, however the electricity flow tends to be from north to south (where population density and load are higher).

Counterintuitive flows occur on a few borders, all involving bidding zones that apply lineset constraints. Note: DK1-NO2, NO1-NO3, NO1-SE3, NO3-NO5 do not have lineset constraints on the border itself, but at least one of the pair of bidding zones involved does apply lineset constraints (on other borders). Counterintuitive flows occur 5-15% of MTUs. As several of these borders consist of HVDC cables, ramping constraints can also cause counterintuitive flows, as explained above for the GR-SUD bidding zone border.

Regarding intuitive flows, borders involving HVDC interconnectors tend to be more congested than AC ones. Beyond this, it is hard to find relevant trends, as borders like SE1-SE2 show price convergence during nearly 100% of MTUs, and borders like NO1-NO3 show congestion ~ 70% of MTUs and ~15% price convergence. Geographically, congestions seem to materialize more often around more densely populated areas (i.e. DK1, DK2, NO1, NO2, SE3, SE4).

All bidding-zone borders with HVDC interconnectors show outages (which are often planned, for maintenance). Note: DE/LU-DK1 and DE/LU-DK2 both have an AC component. However, outages occur on AC borders too (e.g. NO1-NO3, NO1-SE3, NO3-NO5, NO4-SE2).

Price-coupled NTC with allocation constraint - Italy North CCR.

Italy North is, in some ways, a special CCR. The grid itself is rather meshed (17 cross-border connections between NORD and neighbouring bidding zones in the North, of which 9 with CH, 4 with FR, 2 with SI, 2 with AT). However, historically the flow of electricity has been towards Italy, so much so, that the capacity calculation methodology (CCM) of Italy North CCR used to try and maximize imports into Italy explicitly. The “export corner” situation (i.e. NORD exporting on one bidding zone border, and importing on the others) used to be infrequent, while it is now considered in the CCM (and it should go live in Q2 2024). Additionally, Italy North applies allocation constraints on its northern borders.

The combination of these factors results in mostly congested borders (~75-80% of MTUs), with up to 5% of MTUs showing counterintuitive flows. The remaining MTUs are split between price-convergence and outage, with the latter being more common on the AT-NORD border (during 2023 mostly composed of a single network element).

Because the grid in this region is meshed, Italy North CCR and Core CCR will merge into a single CCR (Central Europe) that applies the flow-based method for capacity calculation, following ACER decision 04/2024 of 19/03/2024 (link).

Takeaways

In this third article on counterintuitive flows, I analyzed 2023 data and tried to find trends and patterns. Here are some key takeaways:

• Explicit auctions foster inefficient allocation of cross-zonal capacity and make it very difficult to reach price convergence.
• Regions where the grid is (mostly) radial (SEE, SWE, Baltic, GRIT) typically show good price convergence and almost absent counterintuitive flows.
• Regions that apply flow-based capacity calculation (so far only Core CCR) show a remarkably homogeneous behaviour (excluding Poland) with price convergence during ~1/3 of the MTUs. Counterintuitive flows occur in Core CCR between 10-40% of the time, however in Core commercial schedules are calculated after the clearing prices and net positions are determined.
• Hansa and Nordic CCR show a rather inconsistent behaviour among their bidding zone borders. Generally, borders composed entirely of HVDC interconnectors tend to be congested, and bidding zones applying lineset constraints tend to cause counterintuitive flows on their borders.
• The bidding zone borders part of Italy North CCR are congested during 75-80% of the MTUs

If you would like to discuss these results further, please comment or get in touch via PM (to me or Reinhard K. ).

P.S.: A similar (and more in-depth) analysis of the (in-)efficiency of the EU-GB and EU-CH explicit auctions and counterintuitive flows on these bidding zone borders is in the works, so stay tuned!