Marjan Ilkovski, Boštjan Blažič
Distribution networks are facing major changes as renewable energy sources (RES) and stochastic loads are increasingly being connected. Higher penetration levels of RES, electric vehicles and heat-pumps can lead to operational problems, such as overloading of network elements and power quality problems . Conventional distribution network planning relies solely on network reinforcements, i.e., replacing transformers, changing distribution lines, adding new distribution lines etc., and does not consider the potential impact of load flexibility offered by prosumers in the future . This approach can lead to high investments and oversized networks, where the networks capacity is not fully utilised. To avoid or delay large investments, the evaluation of the potential impact of future load flexibility is necessary. This article will present an approach to distribution network planning based on reference network models.
Network planning with the help of simulations includes three main steps: defining statistically representative network reference models, appropriate modelling of consumption and production and implementation of simulations and, the third step, generalization of simulation results to the entire distribution. Determining the methodology for a statistically sound generalization of the results obtained using the reference models to the entire distribution network is a key step in overall analysis. Such an approach to planning enables both technical and economic comparison of different variants of network development.
Reference models of distribution network
Since it is not expedient (or even feasible) to perform simulations with the entire distribution network model (for example, there are more than 15.000 LV networks in Slovenia), it is necessary to define reference models of distribution network . These models are representative and allow the generalization of simulation results from a limited number of simulation models to the entire distribution network.
Reference models are determined on the basis of limited set of network models and using an appropriate clustering method, which clusters LV and MV networks into groups of networks with similiar characteristics. Each group is represented by reference model, that is a network, that best captures the characteristics of all networks within the group.
The simulation methodology is crucial for the evaluation of the various operational states of users in a particular network. The methodology enables the comparison of the conventional network planning approach, which are typically network reinforcements, with the solutions that additionaly consider the possibility of demand response and voltage regulation. Due to stochastic behaviour and unpredictability of loads, the methodology utilises Monte Carlo method for performing load flow analysis using the developed simulation platform . The basic structure of the simulation platform is shown in Figure 1. LV distribution network model is developed in DIgSILENT PowerFactory using GIS data acquired from DSO. Major part of network modelling process is automated through MATLAB script, which reads the provided network data and constructs an Excel file readable by simulation program. Network customer models include models of household loads, heat pumps, electric vehicles and renewable generation and are used to generate inputs to each simulation. Each model is developed as a Python script and can be upgraded separately. Simulation platform enables the selection of different seasons, weather conditions, year, time of the day, types of the day, number of simulations etc. To consider unsymmetrical conditions, three-phase network model is assumed, allowing the connection of single-phase loads. Demand response algorithm manages consumption of loads when voltage is out of permissible limits or an element in the network is overloaded.
Smart grid solutions included increased demand response of electric vehicles (delaying the charging of electric vehicles to night-time, while ensuring that the batteries are full in the morning), demand response of the heat pumps (heat pumps are turned off in during the peak in the system, but not more than 3 hours per day,while ensuring that the energy not provided in that period is distributed within 3 hours before and after the event), demand response of household consumption, voltage regulation through regulation of reactive power of PV plants (Q(U) characteristic) and voltage regulation with OLTC MV/LV transformer (in a limited number of transformer stations) .
Distribution network planning typically requires prediction of the future state of the network, therefore forecasted number of electric vehicles, heat pumps and amount of PV generation were used to simulate networks operation in the future.
Figure 1: Building blocks of developed simulation platform.
Generalization of simulation results
The second step of network planning includes calculations of power flows and voltages in the reference network models. The final step in distribution network planning is the generalization of simulation results first to the entire group and then to the entire distribution.
With aim of illustrating it, an example of results generalization of the reference model of one group of LV networks is shown .
For every reference network three different states were calculated for winter and summer:
- current state in 2020,
- state in 2030, when heat pumps, electric vehicles and photovoltaic power plants were added, the remaining household consumption was increased,
- state in 2030 by considering demand response algorithm (two scenarios: all users participate, only some users participate).
Based on the sample of LV networks, we determined that the share of networks clustered in the chosen group is 8 %. If we generalize this share to the entire distribution network of Slovenia, we find out that number of networks which belongs to the chosen group is 1246. We found out that in 2030 without taking into account smart grid solutions, 263 transformers will be fully loaded and will need to be replaced with bigger ones. The use of the smart grid solutions in 2030 reduces the number of transformers that are loaded more than 100 % of rated power to 81 in a case when all users participate in demand response and to 153 when only some users participate in demand response.
Also we concluded, that, in the year 2030 without the use of smart grid solutions, at least some of the lines in the group will need to be replaced due to the low voltages. On average, it will be necessary to replace 3800 meters of lines per network or a total of 4735 km of lines.
The high variability of consumption and production, due to the increase of the share of renewable sources and electrification of heating and transport, requires an update of the conventional concept of distribution network planning.
Due to the size of the distribution network, it is advisable to use reference models, ie. a limited set of statistically representative simulation models that allow the generalization of results to the entire distribution. The proposed methodology enables the analysis of a system with stochastic consumption and production and enables us to evaluate the impact of new renewable energy sources and consumers/prosumers on the operation of the distribution system. The methodology also enables assesement of the effectiveness of various solutions, which are necessary steps in network planning.
 L. Gregor, K. Klemen, and B. Boštjan, “METHODOLOGY FOR LOW-VOLTAGE DISTRIBUTION NETWORK PLANNING WITH THE USE OF DEVELOPED SIMULATION PLATFORM,” presented at the 15th Conference of Slovenian Electrical Power Engineers, Laško.
 B. Blažič, G. Lekan, K. Knez, and M. Ilkovski, “Distribution network planning based on reference models,” presented at the 15th Conference of Slovenian Electrical Power Engineers, Laško.
 B. Blažič and et al, “Posodobitev nacionalnega programa pametnih omrežij,” Ljubljana, 2020.