Platform Core

The description of the platform core is based on the methodology deeply explained in [15,72], which is developed by Comillas university. This methodology combined with the precious data collected in the DSO Observatory survey described in chap-ter 2.3, gives the scheme of platform core presented in Figure 3.4.

Figure 3.4: Schematic overview of the methodology used to build the representative distribution network

The first step is the image processing which is done through a Python program that automatically detects buildings in the area under study. The code is avail-able in Appendix C. To minimize the power lines crossing streets and buildings, multipolygon buildings where transformed to a simple polygon through a convex procedure. This allows to reduce the erroneous presence of cables in the courtyard of squared or rectangular buildings. By doing so, a slightly higher number of con-nected customers are assumed, because a bigger surface is estimated. Once the image processing is completed it is sent to the core module box.

The second step is the identification of consumers, which is done by the core module developed by the Instituto de Investigación Tecnológica (IIT) of Comillas Univer-sity in Madrid. At this stage connected customers are assigned to the buildings, and the geo-reference position of the bus name/number is recorded. The distribution of customers is based on the input parameter number and probability of consumers per building % set by the user in the user input phase.

Then, there are two more ”boxes” of information converging toward the green-field RNM step phase as visible in Figure 3.4, and in particular: the RNM catalogue parameter and adjust parameters.

Table 3.1: RNM catalogue for power lines

Type R [Ohms/km] X [Ohms/km] Ampacity [A] Voltage [kV]

overhead 0.65 0.1 150 0.4

overhead 0.42 0.39 250 20

underground 0.14 0.08 420 0.4

underground 0.21 0.11 300 20

First, the RNM catalogue provides the simultaneity factors, the settlement un-derground ratio of both LV and MV lines of the specific area and the network installations. In this latter technical data of LV lines, MV lines, and transformer are extracted. A sample of the RNM catalogue is provided hereby in Tables3.1and 3.2 showing power lines and transformer characteristics. For confidential reasons the investment costs of overhead and underground lines, as well as for transformers are not provided.

The power lines and transformer characteristics were not addressed in the DSO Observatory Survey, but they have been implemented in a second phase through a direct collaboration with DSOs. Therefore, not all countries have specific infor-mation concerning these two elements. In this latter case, default values are utilized.

Table 3.2: RNM catalogue for transformer

Type Capacity [kVA] Secondary voltage [kV] No load losses [kW]

interurban 400 0.4 0.92

interurban 80000 20 47

urban 630 0.4 1

urban 120000 20 71

The Figure 3.5 shown in Mateo et al. [15] pictures the step to identify LV customers (figure on the left), the planning of MV/LV transformer (figure in the centre) and the planning of LV feeders (figure on the right). The same process is then applied for MV customers, which are indicated with bigger black dots, and for the planning of MV feeders coloured in red in Figure 3.12.

It can be observed from Figure 3.5 that the dots, therefore customers, are located at the border of each building. This mechanism has been applied to limit the pos-sibility for power lines to cross edifices. Indeed, cable will follow the dots. The number of dots around each building is mainly defined by the density parameter set by the user in the input procedure. Each dot, is a bus of the network and it has a specific number of ”virtual” connected customer and peak demand which is based on the % of consumer per building.

Downstream to the placement of customers, is the planning of the MV/LV trans-formers, indicated as red dots in the centre of the same figure. The substations capacity (kVA) is interlinked to the peak demand and the layout composition of the urban area. More in detail, if an area is more dense of customer more sub-station will be applied. The core module will select the proper subsub-station from the RNM catalogue, of the DSO operating in that area, based on capacity and cost optimization. Then, LV feeders are constructed to associate LV customers and MV/LV substations.

Figure 3.5: Identification of customer and MV/LV substation positioning The most popular algorithm to develop large-scale distribution networks is the heuristic branch-exchange algorithm. The feeders are constructed based on this algorithm. In our specific case, it is an Euclidean minimum spanning tree problem.

Indeed, the network graph is composed by Euclidean straight edges. Starting from the only known data, the geographical coordinates (x,y) of customers and substa-tions, a minimum spanning tree network is developed. The optimization problem minimize the following cost function, which takes into account fixed (MV network) and variable costs (power losses and reliability):

ψ = ^X

i,j∈E

(K_I+ K_LI_{i j}² )L_ij+ ^X

i,j∈E

K_nI_ijL_j (3.1)

where KI is the coefficient of investment costs, KL and Kn are respectively the cost of power losses and reliability, Iij the current through the branch and Lij the length of the branch.

By having a closer look at the branch-exchange, we start from an initial tree subject to voltage drop and reliability constraints. Then the algorithm through an iterative process select a branch (i,j) and substitutes it with a branch (m,n) that produces a lower value of the cost function written above.

Once the synthetic network is constructed the module recalculates the 10 network indicators⁶.

6DiNeMo automatically detects the geographical area under study and consequently the oper-ating DSO. Each country has different indicators according the data provided by the DSO. In a

The Eucledian Minimum Spanning Tree Problem Branch-exchange technique

A Euclidean spanning tree is a spanning tree of a Euclidean graph, hence it is a circuit-free graph connecting n points in the Euclidean plane. The minimum spanning tree problem becomes the Euclidean minimum spanning tree problem and is: Find the Euclidean spanning tree for which the sum of the Euclidean distances between n points is a minimum [73].

Whether the calculated indicators differ too much from the one provided by the DSO, a new green-field process is run in order to reduce this difference. This con-trol process is also visible in the schematic overview presented in Figure 3.4. The representative network is then developed and provided to the user to be down-loaded from DiNeMo output page (visible in AppendixB). All the relevant outputs provided to the user are explained in the next paragraph.