Portal de Periódicos da CAPES

Integration of open source tools for studying large‐scale distribution networks

2017; Institution of Engineering and Technology; Volume: 11; Issue: 12 Linguagem: Inglês

10.1049/iet-gtd.2016.1560

1751-8695

Gustavo Valverde, Andrés Argüello, Roger V. Gonzalez, Jairo Quirós‐Tortós,

Optimal Power Flow Distribution

IET Generation, Transmission & DistributionVolume 11, Issue 12 p. 3106-3114 ArticleFree Access Integration of open source tools for studying large-scale distribution networks Gustavo Valverde, Corresponding Author Gustavo Valverde gustavo.valverde@ucr.ac.cr Electric Power and Energy Research Laboratory, University of Costa Rica, San José, Costa RicaSearch for more papers by this authorAndrés Arguello, Andrés Arguello Electric Power and Energy Research Laboratory, University of Costa Rica, San José, Costa RicaSearch for more papers by this authorRóger González, Róger González Electric Power and Energy Research Laboratory, University of Costa Rica, San José, Costa RicaSearch for more papers by this authorJairo Quirós-Tortós, Jairo Quirós-Tortós Electric Power and Energy Research Laboratory, University of Costa Rica, San José, Costa RicaSearch for more papers by this author Gustavo Valverde, Corresponding Author Gustavo Valverde gustavo.valverde@ucr.ac.cr Electric Power and Energy Research Laboratory, University of Costa Rica, San José, Costa RicaSearch for more papers by this authorAndrés Arguello, Andrés Arguello Electric Power and Energy Research Laboratory, University of Costa Rica, San José, Costa RicaSearch for more papers by this authorRóger González, Róger González Electric Power and Energy Research Laboratory, University of Costa Rica, San José, Costa RicaSearch for more papers by this authorJairo Quirós-Tortós, Jairo Quirós-Tortós Electric Power and Energy Research Laboratory, University of Costa Rica, San José, Costa RicaSearch for more papers by this author First published: 17 March 2017 https://doi.org/10.1049/iet-gtd.2016.1560Citations: 9AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract The penetration of low carbon technologies (LCTs) is expected to increase in the near future given the attractive incentives from governments, the cost effectiveness of the technologies and the appearance of smart grid schemes. To understand the corresponding benefits and challenges associated with them, power utilities are expected to run more sophisticated studies supported by more accurate and detailed models of distribution network (DN) elements and the network itself. This study presents the development of two software plugins that allow the integration of a geographical information system platform with a distribution system simulator. These plugins are intended to provide power utilities free and open source tools to explore the benefits and impacts of newly adopted LCTs and to analyse smart grid opportunities in large-scale DNs. The effectiveness of the plugins is illustrated considering a real DN with 13,323 customers in Costa Rica. Results demonstrate that the plugins can successfully support in the creation of detailed network models and studies. 1 Introduction Planning studies for distribution networks (DNs) have been fulfilled with the assistance of engineering network models and simulation tools that have helped understanding the network behaviour under particular conditions. At present, most distribution system analysis tools are capable of running unbalanced three-phase power flows and short-circuit analysis, while only a few tools have the capability to perform simulations over a period of time such as day, week, month or year [1]. Modelling of the DN generally ignores the secondary distribution system. Hence, the analysis is usually limited to medium voltage (MV) level, with distribution transformers replaced by aggregated loads. These simplifications were acceptable because low voltage (LV) systems have been considered passive while modelling these secondary systems is not trivial. The increasing integration of low carbon technologies (LCTs), such as distributed energy resources (DER), and the implementation of smart grid schemes at the distribution level brings new modelling requirements in which secondary LV lines and distribution transformers can no longer be neglected, but need to be modelled in detail. As thoroughly discussed in [2], modelling secondary is not easy, and it will take time and effort to accomplish. Having said that, as computer resources grow in power and capability, it is just a matter of time before all utilities have made this important step. Fortunately, most of the information required to model LV elements is readily available in the geographical information systems (GIS) of power utilities [3], hence processing of this GIS data is mandatory. Indeed, GIS provides the needed data management, analysis and awareness to help the smart grid really be smart [4]. Shirek et al. [2] outlines the data recommendations and requirements when performing analyses on secondary systems. In addition, Brenes et al. [5] reports on a database of typical DN element parameters to be used as a complement for GIS data. The authors in [3, 6] report on the creation of hundreds of LV network models from GIS data in the UK. In both cases, the files were exported to Matlab for the data processing and creation of the engineering network models in the corresponding power system simulation tool. However, any change in the GIS data would require a new iteration of data (files) exchange among pieces of software which may be time consuming and even tedious when dealing with long DN feeders and hundreds of secondary systems. Hence, the use of power system tools directly fed from GIS would provide great opportunities for easier handling of simulations [7]. At present time, there are many software tools to carry out DN studies. Although commercial software are well tested and computationally efficient, they do not allow changing the code for adding new algorithms [8], not to mention the license and update costs. Hence, the Electric Power Research Institute (EPRI) and the US Pacific Northwest National Laboratory developed and released the OpenDSS and the GridLab-D, respectively. Two sophisticated open source software tools for the future smart grid studies [9, 10]. An important drawback of these powerful tools is the fact that they do not make direct use of GIS. This is, although they may use the information from GIS as input parameters, these simulators do not interact with the GIS of the power utility. This paper presents the integration of OpenDSS with an open source GIS platform, Quantum GIS (QGIS), to carry out easier and more efficient studies of DNs. The integration is made possible with the creation of software plugins to (i) extract GIS data to automatically build a DN engineering model and (ii) run OpenDSS as an embedded tool in the GIS platform. This integration is a great opportunity to keep an updated and accurate GIS representation of the network. It is expected that a graphical user interface (GUI) for OpenDSS will further enhance the acceptability of OpenDSS, and this opens an opportunity to power utilities and researchers to run their smart grid studies at minimum cost. The plugins will help on planning studies (load growth, network extension, DER integration) but they could also help on operating procedures if some decisions required 'on-line' simulations. In order to provide new opportunities to power utilities, research laboratories and universities to test and simulate novel smart grid solutions, the plugins have been made available in [11]. The rest of this paper is organised as follows. Section 2 gives a brief description of OpenDSS and QGIS and how they are integrated. Section 3 presents a detailed explanation of the network model builder using GIS data while Section 4 presents the type of DN studies that can be carried out in the GIS environment. Furthermore, a demonstration of three network studies is presented in Section 5, followed by a discussion in Section 6 of future complementary tools that will take more advantages of GIS from a smart grid perspective. Finally, Section 7 presents the concluding remarks. 2 Integration of free and open source platforms In the last decade, open source and free software have become more and more popular in the scientific and engineering community. They have the advantage of flexibility, i.e. users can adjust the code, and customise the tools according to particular needs. In addition, they offer universal accessibility to technological advances and make it possible to accelerate software capabilities as user feedback and contributions are regularly incorporated. Indeed, open source software tools have demonstrated to foster further developments, ideas and innovation. In the case of power system analysis, literature [12] provides a list of available open source software tool for a variety of transmission and DN studies. This paper reports on enhanced simulation tools for smart grid studies thanks to the integration and combination of OpenDSS and QGIS. Both software are briefly explained hereafter. 2.1 OpenDSS OpenDSS is an open source software package developed by EPRI that can be used to simulate multi-phase AC DNs [9] for planning and network analysis. It was developed to support the grid modernisation efforts and the integration of DER. A key aspect of this script-driven, frequency-domain simulation tool is that it allows considering the time dimension (e.g. daily simulations with different time step) – critical to quantifying the impacts of variable sources and loads. According to EPRI, the tool was designed to be indefinitely expandable so that it can be easily modified to meet future needs and smart grid studies. To create a network model, OpenDSS follows a sequence of definitions [9] for each network element, i.e. generators, lines, transformers and loads. Hence, given the script-written nature of this software, the creation of large-scale networks must be done carefully, and hopefully, in an automated way. According to EPRI, the software was designed with the recognition that developers would never be able to anticipate what future users will do with it. Hence, a component object model (COM) interface was implemented on the in-process server DLL version of the program. This COM server allows users to integrate OpenDSS with other software packages or programming languages (e.g. Excel with VBA, Matlab, Python) to perform more sophisticated studies. This COM server is the key feature to integrate OpenDSS and QGIS, as explained in Section 4. 2.2 QGIS GIS are computer-based systems used by many service utilities and government institutions to store, read, edit and analyse data referenced by a spatial coordinate system [13]. GIS allow a graphic representation of data. For example, power utilities use GIS to store characteristics and attributes of utility assets such as electric poles, lines, transformers and final customer electric meters, all of them geo-referenced by a coordinate system that allows to find their exact location. The object location is the common attribute among GIS elements. This information is defined according to the appropriate coordinate and projection system. One of the most popular open source tools for geographical systems is Quantum GIS or simply QGIS, under the GNU-GPL license. This free software reads and analyses *.shp files (compatible with commercial software) of GIS layers which contain information of object classes, e.g. layer of MV lines, layer of MV/LV transformers and so on. An important feature of QGIS is that it offers the possibility to develop Python written plugins for extra data processing and analysis not available in the existing software [14]. This is the key feature to integrate QGIS and OpenDSS, as reported in Sections 3 and 4. 3 Engineering DN model builder The first tool reported in this paper is called QGIS2OpenDSS. It was developed to extract and process GIS data of DN elements to generate the OpenDSS files in an automated way. As presented in Fig. 1, this plugin makes use of a GUI that allows the user to select the opened GIS layers to be translated into OpenDSS files. Note that users must define a short name that is later used to define all buses and line segments. Additionally, the user must define the path where the loadshapes are located and the desired location of the resulting OpenDSS files. Fig. 1Open in figure viewerPowerPoint Graphical user interface for QGIS2OpenDSS The tool accepts up to three layers of MV lines, distribution transformers, LV secondary lines, LV services, LV loads and a single substation layer. It is very common to find separate files for overhead lines and underground cables. For underground cables, the user will have to tick the UG option shown in Fig. 1. 3.1 Data requirements This plugin has the following data requirements: HV/MV substation : This layer is optional and should be used when the main power transformer is to be modelled along with the feeder. A single-point layer details the number of windings, impedances, nominal voltages, kVA ratings, number of taps and the short-circuit capacity at HV level. MV lines : The attributes to be found in this layer are the number of phases of the line segment and its identification, i.e. abc for three-phase segments, ab, bc or ca for two-phase branches and a, b or c for single-phase laterals. Additionally, information of neutral and phase conductors such as length, material and conductor size is later used to get the electrical properties of each conductor in a wire database, see [5]. The size can be given in millions of circular mils (MCM), American wire gauge (AWG) or mm2 as all attributes are treated as strings. For overhead lines, the table of attributes must include a line spacing code or letter, e.g. V may refer to vertical line spacing and H for horizontal spacing and so on. The plugin concatenates strings to build a unique OpenDSS line geometry identifier [15]. For example, the identifier 3PMV 477AAC 3/0AAC_H stands for a three-phase MV line in horizontal spacing whose phase conductors are 477 MCM AAC and neutral conductor #3/0 AWG AAC. Underground line segments must also include the insulating material, nominal voltage and cable type (concentric neutral or tape shielded cables). This information is used to extract the electrical properties and dimensions of cables from a database, for details see [5]. Distribution transformers : This layer must include the transformer's nominal voltages and capacity in kVA. For transformer banks, each unit must be specified with a given kVA capacity. Similar to MV lines, it must include the number and identification of phases each transformer is connected to on the MV (primary) side. In the case of three-phase transformers it is imperative to include the winding connection (wye, delta, open wye, open delta). The series impedance and no-load losses of distribution transformers are normally not included in GIS. In order to create an OpenDSS model, the plugin uses a database of typical series impedances based on the nominal voltage and capacity of transformers, see [5]. For single-phase three winding transformers, an approximation is made to calculate each winding impedance based on the transformer's nameplate impedance, as explained in [16, 17]. LV lines : Apart from the length, material and size of neutral and phase conductors, this layer must include the conductor spacing, the number of phases and the operating voltage code to discriminate single-phase three-wire (split phase 120/240 V) from three-phase lines (e.g. 120/208 or 277/480 V). For underground LV cables, the attributes must include the insulation type and cable spacing code. LV wire services : The LV wire service layer must include the length, type, size and material of conductors. These strings are concatenated to define the linecode identifier in OpenDSS [15]. For example, the linecode identifier TRPX2AAC stands for a #2 AWG AAC triplex cable. This identifier is later used to consult a database of electrical and mechanical parameters of service cables to calculate their corresponding primitive impedance matrices, see [5]. LV loads : This layer must include the average monthly kWh consumption and customer type: residential, commercial or industrial in order to allocate load profiles, as explained in Section 3.4. Standardised GIS models are vital to make possible the translation to engineering models. To the authors' knowledge, most power utilities use very similar procedures to represent network elements and store their information in GIS databases. In case that some information is not found in the list of attributes, an error message will warn the user that some attributes are missing. This message may also appear if the GIS layers use attribute names not recognised by the tool. This is easily overcome by changing the corresponding names in the table of attributes as requested by QGIS2OpenDSS. 3.2 Connectivity of elements QGIS2OpenDSS connects two or more elements if their coordinates match or when they are closer than a predefined distance. This is accomplished with Kd-trees. A Kd-tree is a data structure used to organise and manipulate spatial data [18]. This structure allows to quickly find neighbour points within a radius of a query point. The spatial.KDTree class was used to grow the trees in Python. Also the following methods were used: tree.query_pairs(r): It finds all pairs of points within a distance r in the tree. This was used to connect same class elements, e.g. three-phase MV lines. treeA.query_ball_tree(treeB,r): It finds all pairs of points from tree A and tree B whose distance is at most r. This was used to connect different class elements, e.g. three-phase and single-phase MV lines. For all cases, r = 0.1 m. This distance is used to guarantee that small errors in coordinates will not lead to disconnected DN elements. To connect MV lines, the tool builds a tree for each MV line type: single phase, two and three phases. A bus name is created when two line segments ends are found to be connected using the aforementioned packages. In case that one of the segment ends is already assigned with a bus name, the other line segment end will adopt it. For example, the MV bus of transformers adopt the bus name of the MV line segment end they are connected to. In addition, the secondary buses of these transformers are assigned with a new LV bus name. Similarly, in order to find the connectivity of LV elements, the tool builds trees for LV line segment coordinates (both ends), service coordinates (both ends) and load coordinates. Here all LV elements are compared against each other to define all the LV buses. 3.3 Identification of errors The information contained in GIS is very detailed. However, it contains errors that may remain unseen unless a network model needs to be built for engineering analysis. This tool identifies and reports the errors that will affect the simulations in OpenDSS. This report includes the name and coordinates of the problematic elements to facilitate the corresponding correction in the database. After all corrections are carried out, the plugin should be re-run until no errors are reported. The most common error in GIS is the disconnection of DN elements due to small coordinate mismatches in the order of a few centimetres. If any DN element is identified (with Kd-tree) not connected to any other element, the plugin will report it as an isolated element. The tool also checks for erroneous phase designations of MV lines in GIS. The first check is made when two-phase line segments are connected. For example, a bc line segment should never be connected to an ab or ca segment, otherwise an error will be reported. The next check is made when single-phase segments are connected to two-phase line segments. For example, an error is reported if a phase c line segment is connected to an ab line segment. The same procedure is repeated when connecting single-phase segments only. The plugin reports errors when connecting , or segments. Similarly, errors are reported when three-phase transformers are connected to two-phase or single-phase lines, when two-phase transformers are connected to single-phase segment ends, or when the phase identification of a single-phase transformer does not match the phase identification of a single-phase MV line. Finally, the plugin also reports unknown capacities or unknown nominal voltages of distribution transformers. They are considered unknown values when they are not included in the library of transformer impedances. 3.4 Load profile allocation QGIS2OpenDSS allocates a load profile to each customer based on the customer type and monthly energy consumption. The curves must be loaded by the user as shown in Fig. 1. These curves are scaled up or down to perfectly match the customer's monthly energy consumption. These curves are later fine tuned in the load allocation procedure to match the OpenDSS simulation with the feeder's load profile, as explained in Section 4.2. The load profiles are created with a resolution of 10 min, considering different power utilities, type of day (weekday/weekend) and energy consumption range. These profiles were extracted from a statistical analysis of a nationwide measurement campaign in Costa Rica [19]. These load profiles are constantly improved as more measurements become available. It is expected that load profiles from smart meters will replace the ones obtained from the measurement campaign to assess the benefits and impacts of integrating LCTs into the grid [4]. 4 Network analyser The second tool reported in this paper is the QGIS2runOpenDSS. This plugin uses the output files of QGIS2OpenDSS and runs the power system analyser OpenDSS as an embedded tool in QGIS. This is made possible thanks to the COM server that links OpenDSS with Python (see Section 4.1). Fig. 2 presents the GUI in QGIS used to carry out advanced studies for large-scale DNs. The user must select the name of the circuit created by QGIS2OpenDSS in Choose the circuit. The corresponding folder will include a master.dss file which calls the complementary files of lines, transformers and so on. The substation.dss file, if created, is used to automatically load the source voltage, angle and frequency, the short-circuit capacities and the name of the first MV bus connected to the main transformer. Fig. 2Open in figure viewerPowerPoint Graphical user interface for QGIS2RunOpenDSS The user also has to select the feeder's load shape of active and reactive power along with the respective hour and date. In order to run yearly power flows the file must include at least one year of measurements. This information is registered at substation level and stored in a single csv file to carry out the load allocation procedure explained in Section 4.2. The user must define the type of study to be carried out and the folder where the results will be stored. For snapshot studies, it is required to define the simulation date and hour (available in the feeder's load shape file) while yearly power flows require the simulation resolution (e.g. 1, 10, 60 min etc.). The user may also choose to run multiple snapshot or daily simulations if a Monte Carlo approach is to be adopted. 4.1 Integration of QGIS and OpenDSS QGIS drives OpenDSS via the Python console available in the former. The library comtypes.py is used for this purpose, and this is installed using the package management system pip. The following steps are required for this integration: Open the OSGeo4WShell (the command window of QGIS) as administrator. Change the path to find the file 'get-pip.py', which should have been previously downloaded. This is done using the 'cd' command. Install the package management system pip using the following command 'python get-pip.py'. Once installed the pip, execute the following command to install the comtypes : 'pip install comtypes'. It is important to mention that all these steps are automatically executed the very first time the QGIS2runOpenDSS is used (provided that QGIS was executed as administrator), and users are notified about the corresponding installation. 4.2 Load allocation The load allocation procedure tries to match the simulated aggregated demand of all customers and the network losses with the measured demand of the main feeder. This is achieved by iterating and applying a correction factor to all loads (active and reactive power), for each time instant. This iterative procedure is required as network losses are not known beforehand. Let and be the simulated and measured feeder active demand at time instant i, respectively. In the first iteration, i.e. k = 1, the kW correction factor is 1 so that each load is assigned with the original kW value at instant i, taken from the corresponding load profile curve. These demands are used to run an instantaneous (snapshot) power flow simulation to determine the . If the difference between and is larger than a predefined tolerance error, the correction factor is updated. At the iteration, the kW correction factor for instant i is This factor is multiplied to all load values at instant i, and a new power flow simulation is run to obtain the . This iterative procedure is repeated until . Since there are thousands of loads in the circuits, the experience has shown that only few iterative corrections are required to fit the feeder's actual demand, as presented in Section 5.2. A similar procedure is carried out for the reactive power curve fitting. It is expected that information from smart meters will improve the load allocation accuracy to the point that only a small correction will be applied to non-monitored loads and this will also reduce the number of iterations. This load allocation is the first step before running any other network study defined hereafter. 4.3 Types of network studies This section provides some details of the network studies currently carried out in QGIS powered by OpenDSS. The results of these studies are stored according to the user's needs in *.csv files. Snapshot power flows : This is the simplest study to analyse the network conditions during a specific instant (e.g. peak demand and minimum load). This approach, however, might result in under or overestimation of network conditions, particularly when considering the variability of renewable energy sources [e.g. photovoltaic (PV) systems] and loads [e.g. electric vehicles (EVs)]. In order to run a power flow, the user must select the corresponding snapshot box (see Fig. 2) and define the date (dd/mm/yyyy) and time (hh:mm) of the simulation. If the user sets a time that is different to the resolution of the simulation, the plugin will find the nearest available simulation instant. For instance, if the user defines 18:05 h, in a 15 min resolution, the simulation will actually be executed for 18:00 h. Daily power flows : This corresponds to time-series daily power flows. This type of study is very helpful to quantify the impacts of variable sources and loads. The user needs to define the date (dd/mm/yyyy) of the year to be assessed with a given time resolution. For a yearly power flow, the user simply defines the year of interest. Short circuits : In this study, the user must define the bus name and phases to be short circuited. The user can also define a list of bus names and the plugin will create a fault for each bus in OpenDSS. Harmonic power flows : This study is crucial to assess the quality of the service as well as to understand the impact of harmonic distortion produced by loads and sources on DNs. This also allows determining the propagation of current components of frequency other than the fundamental and the resultant distortion of the voltage waveform. If this study is selected, the user must define the number of harmonics to be assessed and the spectrum of each load and source. These studies can be used to assess the impact of high rooftop PV penetration in the DN. For this, the user must enable this option at the bottom of the GUI, as shown in Fig. 2. Here, it is required to define the total capacity to be installed in the circuit and upload a *.csv file with the PV capacity for each customer type and monthly energy consumption level. The list of PV capacities may be the result of a socio-economic study or based on historical data. In the first case, the user knows the most beneficial PV capacity for each customer. In the second case, the user may know the most common PV capacity installed by residential, commercial and industrial customers. This *.csv file is then used to distribute the circuit's total PV capacity among some randomly selected customers. The plugin will search for the customer's monthly energy consumption and allocate the corresponding PV capacity. This allocation stops when the total PV capacity is met. 5 Demonstration 5.1 Circuit builder This section presents a demonstration of the QGIS2OpenDSS tool. It presents the translation of a real 34.5 kV distribution circuit available in the GIS network model of Compañía Nacional de Fuerza y Luz (CNFL), one of the largest power utilities in Costa Rica. This model includes 2168 MV line sections, 7028 LV (secondary) line sections, 512 MV/LV transformers and 13,323 LV loads, all of them stored in different layers. According to the GIS, there are 84 km of MV line sections and 215 km of LV line sections. Fig. 3a presents the GIS model of the circuit section provided by CNFL. The GIS files include all attributes required to r