Geo-Information Is Power: Using Geographical Information Systems to Assess Rooftop Photovoltaics in Costa Rica

Abstract

Economies of scale, incentives, and technological advances have made photovoltaic (PV) systems more affordable and common in developed nations. In Latin America, however, cost and regulation are still barriers for their widespread adoption, particularly by residential and commercial customers. In Costa Rica, there has only recently been interest in rooftop PV systems thanks to a major pilot project carried out from 2010 to 2015 by the Costa Rican Institute of Electricity (Instituto Costarricense de Electricidad, ICE), the administrator of the generation, the transmission system operator, and one of the eight distribution network operators (DNOs) who also supply electricity. The hundreds of installations involved in this project created the momentum needed by the local emerging PV industry to push for changes in the regulatory framework and demonstrate the economic benefits, given the relatively high electricity prices. Although this project represented only a modest PV penetration, ICE recognized that the ability of distribution circuits to host rooftop PV systems should be adequately assessed. This prompted the Ministry of Energy and Environment to create a new and important requirement for Costa Rican DNO s: the generation-hosting capacity of distribution circuits should be quantified considering a comprehensive set of studies, with most of them requiring detailed network models.
Available: https://www.nxtbook.com/nxtbooks/pes/powerenergy_030417/index.php#/p/48

Full text

IEEE Power & Energy Magazine – Issue on DER Integration
Geo-Information is Power
Using GIS to Assess Rooftop PV in Costa Rica
By Jairo Quirós-Tortós, Gustavo Valverde, Andrés Argüello, and Luis(Nando) Ochoa
1. Introduction
Economies of scale, incentives, and technological advances have made photovoltaic (PV)
systems more affordable and common in developed nations. In Latin America, however, cost and
regulation are still barriers for their widespread adoption, particularly by residential and
commercial customers. In Costa Rica, the interest in rooftop PV systems has only recently started
thanks to a major pilot project carried out from 2010 to 2015 by the Costa Rican Institute of
Electricity (Instituto Costarricense de Electricidad, ICE) - the administrator of the generation, the
transmission system operator, and one of the eight Distribution Network Operators (DNOs) who
also supply electricity. The hundreds of installations involved in this project created the
momentum needed by the local emerging PV industry to push for changes in the regulatory
framework and to demonstrate the economic benefits given the relatively high electricity prices.
Moreover, although this project represented only a modest PV penetration, ICE recognized that
the ability of distribution circuits to host rooftop PV systems should be adequately assessed. This
prompted the Ministry of Energy and Environment to create a new and important requirement for
Costa Rican DNOs: the generation hosting capacity of distribution circuits should be quantified
considering a comprehensive set of studies - most of them requiring detailed network models.
The diversity in the installed capacity of rooftop PV systems (from a few to hundreds of kW), as
well as in the voltages they are connected to (from low to medium voltages), means that DNOs in
Costa Rica - and many others around the world - need to have network studies that not only cover
large areas but also adequately model small and medium scale customers. To successfully
achieve this, Costa Rican DNOs have collaborated with the University of Costa Rica in a
pioneering project to develop advanced open-source simulation tools that integrate distribution
network analysis software and multiple Geographical Information Systems (GIS) - from valuable
information of distribution network components and customers stored by DNOs to sun irradiance
and socio-economic statistics.
This article provides an overview of the different aspects to be considered when integrating GIS
data sources and distribution network analysis software. It describes the practical challenges
involving data errors and modelling, the adopted solutions, and, ultimately, the open source tools
developed for the Costa Rican DNOs employing Quantum GIS (QGIS), OpenDSS, and Python.
The use of these tools, created to assess the impacts of PV systems, are illustrated on a real,
large-scale distribution network located in the metropolitan area of Costa Rica.

IEEE Power & Energy Magazine – Issue on DER Integration
2. The Greener the Better
Costa Rica is a country in Central America with almost 5 million inhabitants. Its electricity sector
is characterized by a highly renewable generation mix and a national electricity access above
99% provided by the eight DNOs (see Figure 1 for the geographical extent of these) - all regulated
by the Regulatory Authority of Public Services (ARESEP).
Historically, Costa Rica has been committed to generate electricity from renewable energy
sources, particularly hydro. By 2015, approximately 80% of the total installed capacity (exceeding
3 GW) was from renewable technologies (see Table 1). In terms of energy, 99% of the total
generation (10.6 TWh) was produced from clean energy sources: 75% hydro, 10% wind, 13%
geothermal, 1% fossil fuel and about 1% other technologies (biomass and solar).
For many years, this generation was owned and administered by ICE, complemented by small
private generators and a few utility-scale distributed generators. However, this has changed due
to the increasing penetration of self-generation units installed by residential, commercial and
industrial customers who have found an attractive opportunity to replace some or all the energy
sold by DNOs.
Costa Rica took its first steps in distributed generation back in 2010, when ICE launched a 5-year
pilot project expecting to host up to 10 MW for self-generation units in its license area. The project
offered very attractive conditions, for a fixed period of 15 years, to customers who decided to
install a generating system in their premises. These customers were exempted from installation
costs or access fees, as they were covered by the DNO.
The pilot project was designed to be a living laboratory with the goal of producing commercial and
technical understanding to ICE’s personnel. By the end of the project in 2015, 366 customers
were involved: a total of 3050 kW effectively connected to the grid and 8220 kW awaiting for the
connection permits. Due to the favorable weather conditions in Costa Rica, rooftop PV was the
most attractive technology for self-generation - as it can be seen in Table 2.
This pilot project not only helped the DNO to gain technical and commercial experience but also
strengthened the - at the time - incipient rooftop PV market increasing the number of vendors.
Nevertheless, the penetration level and concentration of PV was so low that it was not possible
to quantify or witness any network impact caused by the new installations.
Although not part of the original plan, the pilot project awakened the interest of customers served
by other DNOs. In particular, the National Company of Power & Light (Compañía Nacional de
Fuerza y Luz, CNFL - a DNO in the metropolitan area of Costa Rica) faced a growing number of
customers requesting permissions to install rooftop PV systems (see Figure 2). Although
commercial conditions were not as good as the ones offered to customers part of ICE‘s pilot
project, the local media and PV vendors played a major role in promoting the technology all over
the country. This pushed ARESEP (the regulator) and the Ministry of Energy and Environment to
create new rules and modify existing regulations to account for new actors in the Costa Rican
electricity scenario: the self-generators.

IEEE Power & Energy Magazine – Issue on DER Integration
Figure 1 License areas of DNOs in Costa Rica
Table 1 Generation installed capacity per technology (2015)
Technology
Installed Capacity (%)
Solar
0.03
Wind
9.07
Hydro
63.09
Biomass
1.30
Geothermal
7.09
Fossils
19.42
Table 2 Results of pilot project carried out by ICE
Technology
No. Customers
Total kW
Solar
360
6758.85
Solar + Wind
2
5.38
Biomass
1
4500
Wind
1
2
Hydro
2
7.5

IEEE Power & Energy Magazine – Issue on DER Integration
Accounting for self-generation was, however, not as smooth and straightforward as initially
expected by the general public and the Costa Rican Association of Solar Energy (ACESOLAR).
It took about three years of consultation processes and regulation amendments before the current
set of rules was finally published by the Ministry of Energy and Environment, along with the grid
access fees defined by ARESEP. These access fees were created as a charge to cover the
network's fixed costs which are shared by all customers.
Figure 2 Location of rooftop PV systems in CNFL (early 2015).
3. Geo-information is Power
The power of GIS lies in its ability to use many types of data related to the same geographical
area, i.e., combining different datasets within a single platform. In the last decade, Costa Rican
DNOs have been using GIS to manage geo-referenced data of network components (e.g.,
location and capacity of transformers; length, spacing and type of conductors) and customers
(e.g., connection point, type, average monthly energy consumption, etc.). For distribution network
analysis purposes, however, this data has only been adopted to manually update Medium Voltage
(MV) network models developed in a separate commercial software package. Low Voltage (LV)
circuits, and, therefore, the corresponding customers, have never been modelled in detail. While
this simplified approach has been effective so far, the assessment of impacts from rooftop PV
systems connected to different voltages requires modelling the circuits in a more sophisticated
way. Not only the LV and MV assets can be simultaneously modelled but customer consumption
and location data can be exploited to produce realistic daily load profiles. The socio-demographics
of the area can also be derived to estimate the potential sizes and locations of rooftop PV systems.
Finally, PV generation profiles can be produced by combining the latter with geo-referenced
irradiance from meteorological databases. However, placing all these large datasets in a single

IEEE Power & Energy Magazine – Issue on DER Integration
GIS platform, and then integrating it with a distribution network analysis software, is not trivial and
requires defining clear procedures.
3.1. Integration of GIS and Network Analysis
A series of steps are needed to successfully integrate GIS and distribution network analysis
software (see Figure 3, the tools will be explained in a later section). As each database comes
with different types of errors, from connectivity to non-standardized labelling and missing data,
the first step is to identify them and, whenever possible, automatically correct them. In general,
engineering criteria is crucial to define an acceptable level of accuracy when fixing these errors.
The next step towards this integration is to combine the data from the corresponding multiple
sources so as to have a single point for data retrieval and thus facilitate exchanges. To create the
electrical models of the circuits, characterize the electricity demand, and produce the PV
generation profiles, the University of Costa Rica used for the project data corresponding to the
network assets and customers (from DNOs), solar irradiance (from the meteorological office) and
socio-economic statistics (from the latest national census). These databases were combined
using QGIS - one of the most popular free and open source tools for geographical systems and
fully compatible with commercial software used by Costa Rican DNOs.
Figure 3 Integration of GIS and network analysis in a single platform
The translation from GIS to electrical models then takes place and depends entirely on the
adopted distribution network analysis software (each of them has its own way to define electrical
models). In this project, the free and open source software OpenDSS - developed by EPRI, USA
- was used to model and analyze distribution networks and selected due to its flexibility and the
possibility to interact with other software through the Component Object Model (COM) interface.
Within this step a simple but important socio-economic assessment is also carried using the
monthly energy consumption and geographical location of customers. This allows creating not
only realistic load profiles but determining whether customers would install rooftop PV systems
(the larger the consumption the more likely). The most adequate size for the PV installation and
the corresponding generation profile can also be determined with this data. The last step in the
integration of GIS and distribution network analysis software is to make QGIS and OpenDSS talk
to each other so as to carry out specific network studies - which can involve from snapshots to
daily and yearly power flows. The open source programming language Python was used to

IEEE Power & Energy Magazine – Issue on DER Integration
integrate QGIS and OpenDSS as well as to drive the tasks associated with each of them. Thanks
to the visualization of GIS software, the corresponding results can then be presented in a way
that is intuitive and therefore easy to understand for technical and non-technical people.
3.2. Practical Challenges
A key aspect in the integration of GIS data is to ensure that all layers use the same projected
coordinate system. It is also critical that the GIS file formats are compatible among themselves -
raster, vector grid and geographic data files are common. However, the most challenging aspect
is that the GIS of DNOs and that of other organizations are prone to human and technical errors
which can be associated with the data itself or the geographical position of objects. Data errors
include components with incorrect or missing attributes, the non-standardized labelling of
components, and the incorrect assignation of phases. For instance, in the GIS data of Costa Rican
DNOs transformers were often found to have capacities equal to 0 kVA and conductors of the
same type to be labelled differently. Three-phase segments were sometimes found to be fed by
a single-phase branch. In general, these and similar data errors can be mostly solved manually
using filters, typical data, and the engineering knowledge of the DNO.
Figure 4 Illustration of typical errors in the GIS data
Errors related to the geographical position of objects bring a different type of challenge: ensuring
that the physical connectivity of any electrical element is reflected in the GIS. For instance,
differences in the coordinates of loads and service cables, transformers and MV lines/cables, and
endpoints of connected segments of lines/cables, are common errors. To illustrate this, Figure 4
shows the disconnection of a load from the service cable, the disconnection of a distribution
transformer from the LV circuit, and the disconnection of two LV line segments. Given the volume
of the data, solving these issues requires not only automatic but sophisticated approaches that

IEEE Power & Energy Magazine – Issue on DER Integration
exploit the topological characteristics of distribution networks. To tackle this, the University of
Costa Rica adopted a series of rules that use kdtree (a data structure to organize and manipulate
spatial data) and graph theory-based algorithms to search for disconnected electrical elements in
the GIS and to reconnect them whenever possible. Errors in the order of millimeters were common
in the analyzed GIS data from Costa Rican DNOs.
Finally, it is important to highlight that the creation of OpenDSS models (or any other software
package) from GIS data requires defining the electrical characteristics of the network components
(e.g., impedances of conductors and transformers). This information, however, is not normally
available in the GIS of Costa Rican DNOs and therefore it needs to be extracted from other
databases (e.g., from typical data or manufacturer datasheets). Depending on the type of network
studies to be carried out, additional information might be needed. For instance, the short circuit
capacity at the substations - critical in fault analyses - were not in the GIS and therefore were
requested to the DNOs.
4. Tools in Action
As previously mentioned, in early 2016 the Ministry of Energy and Environment requested all
DNOs to carry out detailed network studies to identify the distributed generation hosting capacity
of the circuits. This then led to a collaboration between the DNOs and the University of Costa
Rica to develop a platform that integrates the corresponding GIS with distribution network analysis
software. This platform is comprised of the four tools presented as blocks in Figure 3 and
described in detail below. These tools were developed with the free and open source software
QGIS, OpenDSS, and Python, and were designed to assist DNOs with data correction, modelling,
network studies, and visualization of results. Prior to the use of these tools users are required to
combine all network, meteorological, and socio-economic data within one single QGIS project
and enable the COM interface in QGIS. Although the tools are run sequentially, the correction
and modelling tools are necessary only the first time the data is being processed. The network
analysis and visualization tools are run as many times as the required studies.
 Tool 1 - GIS Data Correction: This tool detects common errors in network data and
corrects the ones that can be resolved without user intervention. It reports the exact
location and type of errors that could not be automatically solved so that users can
manually modify the GIS data accordingly. This tool is likely to be needed only when the
network data has not been corrected before - multiple runs will not solve more errors.
 Tool 2 - Translating GIS to Electrical Models: This tool uses advanced algorithms to
connect and produce the corresponding models of distribution transformers, MV and LV
lines, loads, etc. For each customer, the tool creates and assigns a time-series load profile
using the corresponding location and monthly energy consumption. Given that the
demand in Costa Rica changes primarily from weekdays to weekends (seasonal changes
are limited and hence the demand varies little from one month to the other) only one profile
for weekdays and one for weekends is produced. To create these profiles, the tool uses a
statistical characterization of demand derived from power quality monitoring data of more
than 1,000 residential and 400 industrial/commercial customers nationwide collected by
ARESEP in collaboration with the University of Costa Rica. Tool 2 also allows modelling
the existing PV systems - provided a GIS layer with location, capacity and type of panel

IEEE Power & Energy Magazine – Issue on DER Integration
(e.g., standard, thin-film, premium) is available (Costa Rican DNOs are mandated to store
this information). The creation of these files is further explained in “Distribution Network
Model Builder for OpenDSS in Open Source GIS Software”. The input data to this tool are
the corrected GIS layers of the circuit (resulting from Tool 1). This tool is likely to be
needed only once and when the network model has not been created before - multiple
runs will not produce different models.
 Tool 3 - Distribution Network Analyzer: This tool integrates OpenDSS with QGIS.
Crucially, it allows adding more PV systems to the network so as to carry out impact
studies with higher penetration levels from the existing one so eventually the hosting
capacity of the circuit can be quantified. For this, the user introduces the PV penetration
level (in kW) and the tool defines the location and optimal capacity of PV systems based
on an economic assessment. The report recently released by the Ministry of Energy and
Environment details this economic evaluation. The input data of this OpenDSS-based
distribution network analyzer corresponds to the electrical models created using Tool 2,
as well as the active and reactive power demanded by the circuit (information that is also
provided by the local DNO). The latter is used in a load allocation algorithm implemented
as part of this third tool and in which the demand of the circuit is matched with that of the
loads plus losses. This tool can be run as many times as needed to assess each of the
impact studies and PV penetrations of interest. Currently the tool can run snapshot, daily,
yearly and harmonic power flows, as well as short circuit studies. The corresponding
results are stored as layers in QGIS so they can be used by the visualization tool.
 Tool 4 - Visualization of Impacts: The fourth tool enables DNOs to visualize the results
of the impact studies within QGIS, making the identification of lines and buses with
problems easy and efficient. So far, a few different ways to visualize results have been
implemented: color-based ranking of bus voltages and line loading, heatmaps for
identification of areas with voltage and/or line loading problems, animation of results
through a defined period, and element-specific pop-up windows with further results (e.g.,
active and reactive power, loading). This tool can be used immediately after Tool 3 to
visualize the corresponding new results.
To increase the acceptability of the tools by Costa Rican DNOs, a user-friendly graphical interface
was designed for each of them. In addition, the open source nature of the tools makes possible
the implementation of further improvements and/or changes to meet specific requirements of the
DNOs. The tools are currently being used in different types of distribution networks, from rural to
urban, so their effectiveness in solving most of the particularities of Costa Rican networks is
constantly being enhanced.
To illustrate the use of the tools, the large urban network presented in Figure 5 will be used. It
supplies over 5,000 customers (MV and LV) and is located within the concession area of CNFL.
The first tool, which uses as input only the GIS data related to the distribution network, solved all
connectivity issues and reported all transformers with inappropriate capacities (0 kVA) which were
manually changed based on CNFL practices. The corrected data was then combined with the
meteorological (i.e., solar irradiance) and socio-economic data to produce the combined GIS data
which then serves as input to the electrical models.

IEEE Power & Energy Magazine – Issue on DER Integration
The second tool translated the 34.5 kV distribution circuit available in the GIS to an OpenDSS
model. This model contains 1,638 MV buses, 827 MV line sections, 204 MV/LV transformers,
11,174 LV buses, 11,053 LV line sections, and 6,078 LV loads. According to the GIS, there are
24 km of MV line sections and 87.49 km of LV line sections. For each type of day
(weekday/weekend), Tool 2 then created and assigned load profiles to each customer in the
network based on their average monthly energy consumption available in the GIS of Costa Rican
DNOs. Figure 6(a) shows the consumption histogram for this circuit. No PV systems were
connected to this network. Considering a representative weekend in May, Figure 7 illustrates the
total active power profile of the network (case without PV).
Figure 5 Real distribution network used in the illustration
The third tool was then used to run a daily power flow to quantify the impacts from PV systems.
The DNO needs to define the PV penetration and the day to be investigated. For illustration
purposes, a penetration of 1,500 kW is adopted here (approximately 30% of the peak demand).
Tool 3 uses the energy consumption and solar irradiance to assess which customers are more
likely to install a PV system based on economic grounds. It was found that 486 customers (about
8% in the circuit) would have to install a PV systems so as to make the 1,500 kW penetration

IEEE Power & Energy Magazine – Issue on DER Integration
feasible. Figure 6(b) shows the distribution of the corresponding individual capacities. Finally, the
tool uses the chosen day to produce realistic PV profiles for those customers. The resulting total
net active power profile of the network for a weekend in May is shown Figure 7 for a (case with
PV). Tool 3 then runs the daily power flow and finishes by displaying graphically the daily loading
of the main transformer (Figure 7) and all the daily bus voltages (Figure 8) so the DNO can quickly
see the effects of the analyzed PV penetration. In this circuit, the penetration of 1,500 kW in PV
systems reduces the daily energy consumption by approximately 12% (from 79 to 69 MWh) and
the demand at noon by 28% (from 4.5 to 3.24 MW). In terms of voltages, as it can be seen in
Figure 8, some customers experience voltages above the upper limit (1.06 p.u.) around midday -
when PV systems are generating the most. During this period, this figure also shows some
customers with voltages below the lower limit (0.94 p.u.); which is of significant use to the DNO
and highlights that not only voltage rise issues should be addressed.
Finally, Tool 4 is executed to visualize the different problems that might be caused by 1,500 kW
of PV systems in this circuit. A summary of the results is reported via the log message panel in
QGIS showing the number of customers with problems (0.94% in this case) as well as the number
of overloaded lines (4 in MV) and MV/LV transformers (1). Although these impacts can be
considered minor, they represent a bottleneck (i.e., above the hosting capacity) unless the DNO
takes actions to solve the corresponding issues. To further understand the nature and location of
the impacts, DNOs can use the heatmap and color-based ranking features to visualize areas with
problems at specific times of the day. For instance, Figure 9 shows in red the buses with voltages
above 1.06 p.u. at noon (color-based ranking feature) in which a cluster of customers with
problems has been highlighted. The multiple PV installations, some of which with relatively high
capacities (up to 20 kW), considered in that particular LV feeder was the result of the economic
assessment and is aligned with what also happens in different parts of the world: more affluent
customers with high energy consumption are more likely to afford and benefit from PV systems.
This also demonstrates the value to DNOs in combining GIS information to adequately assess
future PV adoption. Indeed, the results will vary significantly in other circuits as they will depend
on the corresponding socio-economic characteristics of the customers.
(a)
(b)
Figure 6 Histograms of (a) energy consumption and (b) defined PV system capacity

IEEE Power & Energy Magazine – Issue on DER Integration
Figure 7 Demand of the circuit without and with 1.5 MW of PV systems (weekend, May)
Figure 8 Daily voltages at customer connection points (weekend, May)

IEEE Power & Energy Magazine – Issue on DER Integration
Figure 9 Buses with voltages above the upper limit at noon (weekend, May)
5. Supporting Future Smart Grids
This article discussed some of the practical challenges when integrating GIS databases from
Costa Rican DNOs and other organizations with distribution network analysis software to assess
the impacts of rooftop PV systems. One key aspect to quantify these effects in a much more
realistic way was the use of socio-economic data (energy consumption and national census data)
when locating and sizing PV systems. In the short to medium terms, the modelling and analysis
tools will be adapted to cater for smart meter data so even more realistic assessments can be
undertaken. This impact analysis tool will soon also incorporate aspects related to power quality
and the coordination of protection devices.
The use of integrated GIS and network analysis software will undoubtedly help DNOs in
adequately assessing the impacts of other low carbon technologies. For instance, to cater for
electric vehicles, DNOs could also use transport data as well as the location of strategic points
(e.g., commercial and leisure areas). Furthermore, solutions can also be explored. Indeed, the
network analyses component can incorporate smart grid schemes that, for instance, involve the
use of smart inverters, the active management of network elements (e.g., on-load tap changers),
electric vehicle management, etc. In fact, DNOs in Costa Rica are now interested in extending
the existing platform to cater for the optimal allocation of fast charging stations for electric vehicles
as well as to investigate solutions to improve the operation of distribution networks.

IEEE Power & Energy Magazine – Issue on DER Integration
The collaboration between Costa Rican DNOs, the Ministry of Energy and Environment and the
University of Costa Rica has demonstrated the tangible benefits that can be achieved when
industry and academia join forces. The positive feedback received from DNOs and the ongoing
improvements are likely to result in a standardized procedure in which Costa Rican DNOs will be
required to use the developed tools to carry out network impact assessments.
6. Further Reading
Bill Meehan, “GIS for enhanced electric utility performance,” Artech House, Boston 2013.
I. J. Ramírez-Rosado, et al., “Powerful planning tools,” IEEE PES Magazine, march/april 2005.
P. Quesada, A. Argüello, J. Quirós-Tortós and G. Valverde “Distribution Network Model Builder
for OpenDSS in Open Source GIS Software” in IEEE PES Transmission and Distribution Latin
America, Morelia. Sept, 2016.
Ministry of Energy and Environment “Techno-economic analysis of distributed generation in CNFL
(in Spanish),” Technical Report, Costa Rica, 2015.
R. González, A. Argüello, G. Valverde and J. Quirós-Tortós “OpenDSS based Distribution
Network Analyzer in Open Source GIS Environment” in IEEE PES Transmission and Distribution
Latin America, Morelia. Sept, 2016.
7. Biographies
Jairo Quirós-Tortós is with the University of Costa Rica, Costa Rica.
Gustavo Valverde is with the University of Costa Rica, Costa Rica.
Andrés Argüello is with the University of Costa Rica, Costa Rica.
Luis(Nando) Ochoa is with The University of Melbourne, Australia, and The University of
Manchester, United Kingdom.