Reduction of the Initial Grid According to Critical Branches Workshop “Enhanced Enhanced pan pan-European European Transmission Planning Methodology” Methodology May 28th & 29th, 2015 S. Lumbreras, F. Banez-Chicharro, L. Olmos, A. Ramos, M. Rivier and J.M. L t Latorre, COMILLAS Th need The d ffor network t k reduction d ti The size and complexity of large-scale power systems make them unmanageable for the purposes of optimal TEP as proposed. It is therefore necessary to reduce the size of the problem problem. STEP 1 – ADEQUACY WITHOUT GRID Scenario and horizon selection STEP 2 – DETECTION OF SYSTEM Snapshot selection Yearly MonteCarlo simulations STEP 3 – NETWORK REDUCTION ACCORDING TO CRITICAL BRANCHES Network reduction Step1: Network partition assigns the nodes to zones STEP 4 – OPTIMAL GRID EXPANSION AT ZONAL LEVEL FROM TODAY TO 2050 STEP 5 – GRID EXPANSION AT NODAL LEVEL Step 2: Calculation of equivalent network parameters STEP 6 – ROBUSTNESS OF THE PROPOSED GRID ARCHITECTURES Methology proposed in eHighway2050, WP8. Reduction of the initial grid according to critical branches. 2 S Summary description d i ti off th the ttask k Input : - Nodal network description - Indices from power flow simulations/snapshots p Identification of critical branches Several indicators are calculated in order to identify the lines that have the most critical impact on the operation of the system Network partition Nodes are clustered into zones Reduced network calculation Calculation of zonal equivalent network Output : - Assignment of nodes into zones - Zonal network description Reduction of the initial grid according to critical branches. 3 C it i ffor network Criteria t k partition titi Most of the references that deal with network partition use electrical distance to guide the process. The electrical distance between a pair of nodes is defined as the equivalent impedance between them ii.e., them, e the voltage drop between the nodes when a current of 1 A is injected in one and withdrawn from the other. This equivalent impedance is computed using the elements of the inverse of the admittance matrix (i e Zbus): (i.e. Dij = Zbusii +Zbus jj -2Zbusij Zbus = Ybus -1 ( Ybusij = - Rij +jX ij Ybusii = å (R ) +jX ij ij j -1 ) +å jB -1 ij ,i j Pairs of nodes that are connected by short, low reactance lines present lower distances between them. [1] Haixia Wang; Rao Liu; Weidong Li; Caihong Zhao, "Power Flow Tracing with Consideration of the Electrical Distance," Power and Energy Engineering C f Conference, 2009 APPEEC 2009. 2009. 2009 Asia-Pacific A i P ifi , vol., l no., pp.1,4, 1 4 27-31 27 31 March M h 2009 Reduction of the initial grid according to critical branches. 4 C it i ffor network Criteria t k partition titi (cont’ed) ( t’ d) Several techniques have been applied to network partition in the literature. Clustering algorithms such as K-means or K-medoids [2]: efficient, easy to implement. Other approaches: MIP (unmanageable), Spectral Clustering [3] (useful in some cases but difficult to implement) If network partition is established solely on the basis of electrical distance, all considerations other than equivalent impedance are disregarded: system operation, demand and generation, line capacities and congestion. There is no guarantee that frequently congested lines will not belong to the same partition, so they will not be represented. This might lead the planning process to miss potentially important reinforcements. reinforcements [2] E. Cotilla-Sanchez, P. D. H. Hines, C. Barrows, S. Blumsack and M. Patel, "Multi-Attribute Partitioning of Power Networks Based on Electrical Distance," Power Systems, IEEE Transactions on, vol. 28 (4), pp. 4979-4987, 2013. [3] C. Hamon, E. Shayesteh, M. Amelin and L. Söder, "Two partitioning methods for multi-area studies in large power systems," International Transactions on Electrical Energy gy Systems, y , 25 ((4), ), 648-660,, April p 2015 Reduction of the initial grid according to critical branches. 5 C it i ffor network Criteria t k partition titi (cont’ed) ( t’ d) We propose to build the partitions around transmission lines that are particularly relevant for the purposes of TEP. We will refer to them as critical branches. Frequent congestions Power flow control devices (HVDC lines, PSTs). We propose a heuristic algorithm that builds an initial partition around the critical branches. We propose to base the partition on a composite distance (electrical + geographical) in order to avoid overlapping partitions partitions. Reduction of the initial grid according to critical branches. 6 G General l idea id off th the ttestt case study t d Number of nodes ൎ 1,000 400 kV nodes = 548 ൎ 3,715 Number of circuits ൎ 3 715 400 kV lines = 981 PST = 8 Reduction of the initial grid according to critical branches. 7 Id tifi ti off critical Identification iti l branches b h The output of Monte Carlo (MC) simulations is used to calculate a range of indicators that are used to identify critical branches: Average flow Proportion of simulations/snapshots where a line is congested Congestion severity, calculated as the dual variable of flow constraints (marginal severity) or as nodal price differences differences. The correlation among these congestion severity indicators is relatively low. Relationship among indicators in the test case study (average flow (pu), congested snapshots (h),among marginal severity (€/MW) nodal price Severity, difference ) Relationship the indicators (Average Flow, Congestion, Price(€/MW) Diff) 1 0.5 0 6000 4000 2000 0 2000 1000 Congestion severity indicators (marginal severity and nodal price difference) are the only highly correlated indicators. 0 2000 1000 0 (based on 3714 lines, 8760h) 0 0.5 1 0 200040006000 0 1000 2000 0 1000 2000 Reduction of the initial grid according to critical branches. 8 Id tifi ti off critical Identification iti l branches b h ((cont’ed) t’ d) However, independently of the specific indicator choice, critical branches tend to appear in the same areas. areas Impact of the selection criterion on the identification of critical branches (in blue) in the test case study Critical branches according to average flow Critical branches according to congestion Average Flow Congestion x 10 6 6 x 10 2.8 2.8 2.6 2.6 2.4 2.4 2.2 2.2 2 2 1.8 1.8 1.6 1.6 1.4 1.4 1.2 1.2 1 1 -4 -2 0 2 4 6 8 10 12 14 -4 -2 0 2 4 6 8 10 12 5 x 10 Marginal severity 6 Nodal price difference 6 Critical branches according to marginal severity x 10 Critical branches according to price differences x 10 2.8 2.8 2.6 2.6 2.4 2.4 2.2 2.2 2 2 1.8 1.8 1.6 1.6 1.4 1.4 1.2 1.2 1 14 5 x 10 1 -4 -2 0 2 4 6 8 10 12 14 -4 5 x 10 -2 0 2 4 6 8 10 12 14 5 x 10 Reduction of the initial grid according to critical branches. 9 N t Network k partition titi Aggregates the nodes of the network into areas. Several considerations are taken into account: Electrical distance Geographical distance Electrical distance (y-axis, in ohm pu) 0.06 vs. geographical distance (x-axis, in m) Distribution of electrical distance in theDistribution test ofcase study electrical distance 1800 0.05 1600 1400 0.04 1200 0.03 0 03 1000 800 0.02 600 400 0.01 200 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 x 10 Electrical and geographical distances are relatively well correlated in the test case study. 5 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Outliers in this distribution represent nodes that are relatively isolated from the rest of the network. They are eliminated from the network partitioning process in order to avoid the definition of single-node single node zones. zones Reduction of the initial grid according to critical branches. 10 N t Network k partition titi (cont’ed) ( t’ d) A heuristic has been developed to generate a relatively good starting partition partition. For each critical branch, if both extremes of the line belong to the same area, then the area is split in two and nodes are reassigned to the two new partitions based on distance. A purpose-built clustering algorithm, based on K-medoids, takes into account critical branches at every step of the algorithm: The algorithm works iteratively to find a local optimal clustering solution (there is no guarantee of global optimality): 1. 2. Nodes are assigned to the closest medoid. Medoids are recalculated for each area as the existing node that presents the least average distance to the nodes of that area. Nodes that belong to the extremes of a critical branch can never be clustered together. Reduction of the initial grid according to critical branches. 11 N t Network k partition titi (cont’ed) ( t’ d) Network partition has been performed for several definitions of aggregate distance: 50% Geographical / 50% electrical distance In order to use electrical distance, only HV nodes (400kV) can be considered, once relatively isolated substations have been removed (the remaining nodes will be assigned according to the nearest 400kV node to which they have some connectivity). The incorporation of geographical distance avoids the definition of geographically overlapping partitions. The representative node in each partition can be selected as the most central one (lowest average distance to the remaining nodes in the zone) or as the most connected node, which leads to selecting the most significant nodes in the network (large power plants or substations). Several clustering algorithms have been tested: Purpose-built version of K-Medoids that uses the developed initial partition as a starting solution l ti and d checks h k every node d assignation i ti tto ensure th thatt critical iti l b branches h are respected. t d K-means and fuzzy C-means that are heuristically modified to incorporate critical branches. The solutions are relatively y similar for all the approaches pp tested. Reduction of the initial grid according to critical branches. 12 T t results Test lt Initial partition found by the heuristic 24 zones are needed to respect the critical branches 6 x 10 2.6 2.4 2.2 2 1.8 1.6 1.4 4 1.2 1 -4 4 -2 2 0 2 4 6 8 10 5 x 10 Reduction of the initial grid according to critical branches. 13 T t results Test lt (cont’ed) ( t’ d) Network partition based on 100% electrical distance (50 partitions) Some partitions have a geographical overlap 6 x 10 2.6 2.4 2.2 2 1.8 1.6 1.4 4 1.2 1 -4 4 -2 2 0 2 4 6 8 10 5 x 10 Reduction of the initial grid according to critical branches. 14 T t results Test lt (cont’ed) ( t’ d) Network partition based on 50% electrical / 50% geographical distance (50 partitions) Geographically overlapping partitions are avoided 6 x 10 2.6 2.4 2.2 2 1.8 1.6 1.4 4 1.2 1 -4 4 -2 2 0 2 4 6 8 10 5 x 10 Reduction of the initial grid according to critical branches. 15 T t results Test lt (cont’ed) ( t’ d) If the most central nodes are selected as medoids, medoids the representative nodes of each zone are not necessarily nodes that are representative of the network. 6 x 10 2.8 Medoids map Medoids selected by centrality (low distance) 2.6 2.4 2.2 2 1.8 1.6 14 1.4 1.2 1 -4 -2 0 2 4 6 8 10 12 5 x 10 Reduction of the initial grid according to critical branches. 16 T t results Test lt (cont’ed) ( t’ d) If the most connected nodes are selected as medoids (post-processing), (post processing) the representative nodes of each zone correspond to the most significant nodes in the network. 6 Medoids map x 10 Medoids selected by connectivity Connectivity map 2.8 2.6 2.6 2.4 24 2.4 2.2 2.2 2 2 18 1.8 1.8 1.6 1.6 1.4 1.4 1.2 1.2 1 1 -4 -2 0 2 4 6 8 10 -4 -2 0 2 4 6 8 10 12 5 x 10 Reduction of the initial grid according to critical branches. 17 E i l t reduced Equivalent d d network t k Obtain reactances and capacities of zonal (reduced) network. The equivalent network is unique for all the snapshots. The inter-zonal network is represented by equivalent corridors Both critical and non non-critical critical branches. branches But special lines must be represented explicitly Idea: inter-zonal flows of the zonal network “must” match inter-zonal flows in the nodal network. Reduction of the initial grid according to critical branches. 18 R Reactances t off equivalent i l t corridors id Inter-zonal corridors are the equivalents of the set of lines connecting two zones in the nodal network model. model The reactances of the inter-zonal corridors are calculated so that the differences in inter-zonal flows between the existing nodal network and the equivalent zonal network are minimized. Reactances that minimize the Euclidean norm of the difference between inter inter-zonal zonal power flows. A single reactance is obtained for all snapshots. Shi, D., and Tylavsky, D.J. “An improved bus aggregation technique for generating network equivalents” IEEE Power and Energy Society General Meeting, July 2012 10.1109/PESGM.2012.6344668 Reduction of the initial grid according to critical branches. 19 R Reactances t off equivalent i l t corridors id Inter-zonal corridors are the equivalents of the set of lines connecting two zones in the nodal network model. model The reactances of the inter-zonal corridors are calculated so that the differences in inter-zonal flows between the existing nodal network and the equivalent zonal network are minimized. Reactances that minimize the Euclidean norm of the difference between inter inter-zonal zonal power flows. A single reactance is obtained for all snapshots. Equivalent reactances Shi, D., and Tylavsky, D.J. “An improved bus aggregation technique for generating network equivalents” IEEE Power and Energy Society General Meeting, July 2012 10.1109/PESGM.2012.6344668 Reduction of the initial grid according to critical branches. 20 R Reactances t off equivalent i l t corridors id Inter-zonal corridors are the equivalents of the set of lines connecting two zones in the nodal network model. model Example: 9 nodes – 3 zones R Reactances t off equivalent i l t corridors id Inter-zonal corridors are the equivalents of the set of lines connecting two zones in the nodal network model. model Example: 9 nodes – 3 zones Zone A Zone B Zone C R Reactances t off equivalent i l t corridors id Inter-zonal corridors are the equivalents of the set of lines connecting two zones in the nodal network model. model Example: 9 nodes – 3 zones Z Zone A Zone A X=0.0172 Zone B Zone B X=0.0760 X=0.0767 X=0.04 Zone C Flow A-B Fl A B = -41 41 Flow A-C = 16 Flow B-C = -25 These two lines may have different characteristics Zone C Flow A Fl A-B B = -41 41 Flow A-C = 16 Flow B-C = -25 Reduction of the initial grid according to critical branches. 23 C Capacity it off equivalent i l t corridors id Inter-zonal corridors are the equivalents of the set of lines connecting two zones in the nodal network model. model Lack of references: usually aggregation of capacity The capacities of the inter-zonal corridors are calculated so that they allow the maximum interinter zonal flows that could happen in the nodal network Idea Nodal network Maximum power transfer among zones Zonal network Warning: overestimate/underestimate capacities among zones Flows Equivalent capacities Reduction of the initial grid according to critical branches. 24 C Capacity it off equivalent i l t corridors id Inter-zonal corridors are the equivalents of the set of lines connecting two zones in the nodal network model. model The capacities of the inter-zonal corridors are calculated so that they allow the maximum interzonal flows that could happen in the nodal network. 1. Maximum power transfer among zones in the nodal network Maximum power that may be exchanged among zones Use several operation situations: snapshots Look for maximum inter-zonal flows Diversity of situations Zone C Zone A Z Zone B Reduction of the initial grid according to critical branches. 25 C Capacity it off equivalent i l t corridors id Zone C Zone A Gen Zone C Zone A Zone B Dem Gen Zone B Dem Reduction of the initial grid according to critical branches. 26 C Capacity it off equivalent i l t corridors id Inter-zonal corridors are the equivalents of the set of lines connecting two zones in the nodal network model. model The capacities of the inter-zonal corridors are calculated so that they allow the maximum interzonal flows that could happen in the nodal network 1. Maximum power transfer among zones in the nodal network Two different methods: 1) Calculated among zones connected by corridors For each snapshot Repeated for every two pair of zones and in both directions One zone is exporting and the other is importing (the other way is also considered) • Increase generation in exporting and increase demand in importing homothetically Zone C Zone A Gen Zone B Dem Obtain a set of new power injections Reduction of the initial grid according to critical branches. 27 C Capacity it off equivalent i l t corridors id Inter-zonal corridors are the equivalents of the set of lines connecting two zones in the nodal network model. model The capacities of the inter-zonal corridors are calculated so that they allow the maximum interzonal flows that could happen in the nodal network 1. Maximum power transfer among zones in the nodal network Two different methods: 1) Calculated among zones connected by corridors For each snapshot Repeated for every two pair of zones and in both directions One zone is exporting and the other is importing (the other way is also considered) • Increase generation in exporting and increase demand in importing homothetically Zone C Zone A Dem Zone B Gen Obtain a set of new power injections Reduction of the initial grid according to critical branches. 28 C Capacity it off equivalent i l t corridors id Inter-zonal corridors are the equivalents of the set of lines connecting two zones in the nodal network model. model The capacities of the inter-zonal corridors are calculated so that they allow the maximum interzonal flows that could happen in the nodal network 1. Maximum power transfer among zones in the nodal network Two different methods: 2) Calculated for the whole system Only once for each snapshot Now exporting and importing nodes, not zones • Increase generation in exporting and increase demand in importing homothetically Dem Zone C Zone A Gen Gen Zone B Dem Gen Obtain a set of new power injections Dem Reduction of the initial grid according to critical branches. 29 C Capacity it off equivalent i l t corridors id Maximum transfer calculated before: nett sett off power injections i j ti Reduction of the initial grid according to critical branches. 30 C Capacity it off equivalent i l t corridors id Inter-zonal corridors are the equivalents of the set of lines connecting two zones in the nodal network model. model Example: 9 nodes – 3 zones Z Zone A Zone A Zone B X=0.0172 Zone B X=0.0760 X=0.0767 NTC=500 These ttwo o lines ma may ha have e different characteristics Zone C Capacity Among zones Whole system Aggregate (usual) A-B 88 69 100 A-C 167 81 1000 BC B-C 175 72 500 Zone C Intra-zonal constraints affect inter-zonal corridor capacities Reduction of the initial grid according to critical branches. 31 C Capacity it off equivalent i l t corridors id Inter-zonal corridors are the equivalents of the set of lines connecting two zones in the nodal network model. model Example: 9 nodes – 3 zones Z Zone A Zone A Zone B X=0.0172 Zone B X=0.0760 X=0.0767 NTC=500 These ttwo o lines ma may ha have e different characteristics Zone C Capacity Among zones Whole system Aggregate (usual) A-B 88 69 100 A-C 167 81 1000 BC B-C 175 72 500 Zone C Explore the option that intra-zonal constraints do not affect inter-zonal NTC Reduction of the initial grid according to critical branches. 32 C Case study: t d d description i ti 2012 French-Spanish HV system Neighbor countries modeled as a reduced set of nodes Number of nodes 2,196 Number of external nodes (no FR-ES) 211 Number of 400kV nodes 548 Number of lines 3,715 Number of 400kV lines 981 Number of 400kV PST 8 Number of 400kV interconnexion lines 45 Number of HVDC lines 0 Reduction of the initial grid according to critical branches. 33 C Case study: t d d description i ti 2012 French-Spanish HV system Neighbor countries modeled as a reduced set of nodes Two different scenarios are built for 2050 Values of installed capacities are interpolated for 2030 and 2040: three time horizons France Scenario 1 Scenario 2 Spain Decrease nuclear g generation End of nuclear g generation by y 2050 Increase wind and solar capacities by a factor 3 by 2050 Increase wind and solar capacities Constant load Increase load by 50% by 2050 Increase hydro generation No change in hydro Increase gas and coal generation No change in thermal Increase wind and solar capacities by 50% by 2050 Increase wind and solar capacities by 50% by 2050 Increase load by 30% by 2050 Constant load High increase in SP demand compensated with wind and solar in FR Increase demand in FR For each scenario and time horizon 10 Monte-Carlo years are generated Reduction of the initial grid according to critical branches. 34