Challenge: How to define a model of the pan-European transmission system at 2050 which is an optimal trade-off between accuracy of data and time of calculations of grid simulations studies? What would be an appropriate number of nodes for such simulations?
Background and assumptions
In short term studies, system simulations as well as load flow analysis at nodal level are required to define the grid needs. However, the whole European transmission network includes almost 10.000 electrical nodes (Figure 1), which are not tractable for the project. Indeed, for such a long term horizon, uncertainties increase when we perform geographical zooms. A clustering approach has thus been introduced, reducing the level of description of the grid.
Figure 1 : present European transmission grid
Description of the result
The geographical clustering process is performed to split Europe and its countries into smaller parts, relevant for the system modeling (i.e. they should not be too big, to also represent grid characteristics but large enough to enable a geographical allocation of generation and demand in 2050).
The basis for this analysis has been the Nomenclature of Territorial Units for Statistics (NUTS3 regions) which is set by Eurostat.
The clustering is also based on real system characteristics in order to represent the underlying transmission system appropriately. The clusters must be valid for all scenarios of the project.
The criteria used are population, potential of RES generation, land use – availability of the areas, installed generation capacity (thermal and hydro).
A two-step approach was implemented. In a first step an algorithm (K-means clustering) is applied considering the first four criteria which are all measureable criteria. The optimization aims at joining together the incremental NUTS-regions in a way that homogenous clusters are reached.
Figure 2 : NUTS regions
In the second step, the TSO expertise is taken into account for the optimization of the clusters.
Then, the clusters are used for the definition of transmission equivalents. The starting point is the existing transmission system, including the grid reinforcements already planned for the next decade. This whole picture is provided by the Ten Year Network Development Plan (TYNDP) . This basis for grid reduction ensured that the European grid model is based on the most accurate information about the available transmission system.
The determination of transmission equivalents, enabling scheduled unit commitment optimization and grid analysis, is heading towards two main indicators. A transmission equivalent is characterized by its thermal capacity and impedance, latter describing the load flow distribution within the network. Equivalent lines were only introduced between adjacent clusters sharing at least one interconnection line in reality.
The methodology chosen to determine the thermal capacity of the transmission equivalent has been derived from European Network TSO-E (ENTSO-E) methodology to assess the Grid Transfer Capacity (GTC) value between two neighboring countries.
The purpose of Z-equivalents in the grid model is to estimate load flows of the reduced system in comparison to real flows on the borders between clusters. The used methodology searched for and optimal impedance matrix, that minimizes the mean to Root Mean Square Error (RMSE) on the difference between initial flows of the nodal and the reduced network, for each transmission equivalent.
The final pan-European cluster model, as a result of the clustering and grid reduction, is shown in Figure 3, with almost 100 clusters.
Figure 3 : Clusters of Europe
Assessment of the methodology use and limitations
The European grid model is used by the e-Highway2050 project as a tractable model to perform network simulations with respect to identify candidate grid architectures able to meet challenges of electricity markets until the 2050 time horizon.
Beyond the e-Highway2050 project, such clustering approach could be used in further grid planning studies at national or EU levels.
This article is connected to the following e-Highway2050 knowledge articles:
"Energy Roadmap 2050 Impact assessment and scenario analysis,” European Commission
G.Oettinger, EU Commissioner for Energy,“A pan-European grid for 2020 and beyond”
“Guidelines for trans-European energy infrastructure”, European Commission, October 2011
B.H.Bakken, M.Paun, R.Pestana, G.Sanchis, “e-Highway2050: A Modular Development Plan on Pan-European Electricity Highways System for 2050”, Cigre Lisbon, April 2013
e-Highway2050 web site
- ENTSO-E, “Ten-Year Network Development Plan (TYNDP)”, 2012
- Brahim Betraoui, Gerald Sanchis, Nathalie Grisey, RTE. Address: Tour Initiale, 1 Terrasse Bellini, 92800 Puteaux, France. Phone: +33 1 39 24 41 24. E-mails: email@example.com; firstname.lastname@example.org; email@example.com
- Thomas Anderski, Yvonne Surmann, Amprion GmbH. Address: Rheinlanddamm 24, 44139 Dortmund, Germany. Phone: +49 231 5849 16529. E-mail: Thomas.Anderski@amprion.net; Yvonne.Surmann@amprion.net
 k-means clustering is a method of vector quantization used for cluster analysis in data mining. k-means clustering aims to partition n observations into k clusters in which each observation belongs to the cluster with the nearest mean, serving as a prototype of the cluster.
 ENTSO-E, “Ten-Year Network Development Plan (TYNDP)”, 2012