Urban-Scale Rooftop Photovoltaic Potential Estimation Using Open-Source Software and Public GIS Datasets ^†

Highlights

This study highlights the needs and means of estimating open-source photovoltaic power output in future urban areas. The proposed methodology includes the building shadow analysis as the first of the key inputs. The second key input is the statistical analysis based on the PVGIS tool, which consists of 16-year data processing and the evaluation of the photovoltaic installation aspect, slope, and, therefore, general efficiency of such systems in the specified sample area in Bratislava, Slovakia. The results from the proposed methodology are presented, and future research in open science in this area is discussed.

What are the main findings?

• The photovoltaic power estimation in predominantly urban areas without building shadow analysis is not a good approximation, especially in winter, when the building shadows cover the largest areas. This aspect is often neglected in studies with very large sample areas.
• Many current methodologies for photovoltaic power output estimation are of a proprietary nature, resulting in limited use in academic, public, or other non-commercial use.

What is the implication of the main finding?

• Urban areas need to be examined with proper tools and algorithms, with the most precise knowledge of meteorological, geospatial, and geographical data.
• There is a strong need for a completely open-source methodology for various aspects of urban planning, including smart city concepts, energy efficiency, and low-carbon footprint engineering in general.

Abstract

This paper aims to effectively estimate urban-scale rooftop photovoltaic potential using strictly open-source software and publicly available GIS data. This approach is often neglected; however, its importance is significant regarding technology transfer and general commercial or academic ease of use. A complete methodology is introduced, including the building shadow analysis. Although many papers are published in similar areas, very few reveal the specific steps and functions in the software used, or the computational core of some part of the estimation is a “black box” of a commercial service. Detailed irradiation parameters can be obtained using the proposed methodologies, and the maximum photovoltaic (PV) power output in the area can be estimated. The great advantage of this model is its scalability and the easy way of modifying every computational parameter. The results and limitations of the proposed methodology are discussed, and further development is suggested. The presented model is based on a sample location in Bratislava, Slovakia, with an area of circa 2.5 km².

1. Introduction

To address the future needs of climate change and carbon footprint mitigation, urban areas are continuously facing new challenges, mainly regarding energy efficiency and environmental questions. These measures are primarily aimed at improving the quality of life of cities’ inhabitants and contributing to global efforts to eliminate greenhouse gasses. That being said, the best examples of such efforts are banning diesel cars in favor of cleaner and highly efficient electric vehicles [1] in some major European cities like Stockholm [2] and adopting new urban energy management concepts, such as distributed generation and smart grids, which can improve energy efficiency and sustainability [3]. Effective urban planning needs multiple data sources, including the energy demand, energy supply, or building potential, along with tools to assess the outcomes of the intervention [4]. The proposed tool for rooftop photovoltaic (PV) estimation fits such urbanistic studies perfectly.

These honorable efforts are primarily made and initiated by the European Commission. They are valid only for the European Union (EU) member states, even though their CO₂ emissions are around 7% of total CO₂ emissions produced in the world [5]. As can be seen from Figure 1, the European Union is, thanks to these new standards, lowering CO₂ emissions production over the years. That, indeed, cannot be said for most of the world.

To mention some of the most important of the initiatives, there is a provisional agreement between the European Parliament and the Council to strengthen the EU Renewable Energy Directive, marking progress toward the “Fit for 55” [6] legislation to achieve the European Green Deal and REPowerEU objectives. The Fit for 55 package, aimed at aligning the EU with its decarbonization goals by 2050, has led to proposed revisions to key energy directives. This includes the Energy Performance of Buildings Directive and the Renewable Energy Directive. The former [7] requires that buildings meet minimum energy performance standards, focusing on renovating existing structures. The latter directive aims to increase the share of energy from renewable sources, reflecting a crucial step in achieving decarbonization targets.

Current Trends in Research

Identifying all possible locations to expedite the progress toward decarbonization and fully utilize solar potential is a significant step. Data for research and commercial purposes can be obtained from public repositories, online geoportals, or volunteer geographic-information-based websites like OpenStreetMap (OSM). Unlike traditional maps, which are prone to generalization, these archives are translated into GeoDatabases, which store data with potentially very high accuracy. Future PV installations are expected to expand significantly to urban areas and rooftops. Newly available data with high precision are beneficial for future urban PV systems estimations in that manner. In a broad view, this is good news for future research since LiDAR and similar classified point cloud data (mostly in .LAS format), DSM (Digital Surface Model), and DTM (Digital Terrain Model) are more and more common public resources. Today’s research in this area is therefore understandably oriented toward similar goals as this paper.

However, many current studies are extremely complex in terms of the specific 3D roof modeling [8], or complex deep learning algorithms [9,10]. Other similar studies use commercial software tools [11] with a more hardware-heavy oriented approach, for example, using massive point cloud data source files. Some studies try to directly connect the PV power output to the power grid simulations [12], but the simulations itself are more of a demostrative nature, not pointing toward specific implications in the grid. The area examined in this way is usually very limited—this is understandable because of the complexity of the problem. Also, the simulation software is not in open-source licensing too. The estimation of individual building sustainability of large regions by using PV installations is attempted by [13], which is a very complex study and not suitable as a tool for PV power prediction, while also not considering any shadow analysis. Similar studies are conducted in the areas of Taiwan [14], Shanghai [15], Sri Lanka [16], Germany [17], part of the US [18], or all of the European Union [19]. Each of these studies is missing either shadow analysis and/or the targets set for this paper: fully open-source availability of software and source gaospatial data, as simple a tool as possible, and as lightweight a tool as possible. All the mentioned papers indicated that this research area is highly topical for current socio-economical and engineering needs.

Our proposed model, attempting to fulfill all the mentioned targets, was tested on the urban area of Bratislava, with a size of 1948 × 1287 m, which is roughly 2.5 km². This paper is part of a continuous research effort on this topic from previous authors’ publications such as [20,21,22].

Multiple platforms are currently available for PV power output estimation. Some systems are paid services and offer more complex analyses, while open-source alternatives are only a part of the solution. Table 1 summarizes the currently most important applications and services in the fields of PV power estimation. Our proposed solution is different, mainly from the perspective of the open-source licensing, building shadow analysis, and the statistical evaluation of typical 1 kWp installation. This is described further in the text in more detail. The structure and content of the following sections of the article are as follows.

Table 1. Overview of solar rooftop assessment tools.

Tool Name	Description	Availability
ArcGIS Solar Radiation toolset	These tools assess the suitability of rooftops for solar PV installations and incorporate GIS data layers such as building footprints, terrain models, and tree canopy data.	Paid software [23]
Photovoltaic Geographical Information System (PVGIS)	PVGIS is an online tool developed by the European Commission’s Joint Research Centre. It provides worldwide high-resolution solar radiation and PV performance data, allowing users to estimate energy yield of rooftop PV systems based on location-specific parameters.	Free software [24]
SOLARGIS	Developed by GeoModel Solar, the same researchers behind PVGIS, it offers extensive studies on solar irradiation, PV power estimation, and PV investment evaluation. It is a commercial company with numerous scientific outputs.	Paid software and services [25]
Project Sunroof	A web-based tool developed by Google using aerial imagery and 3D modeling to estimate solar potential of rooftops in specific geographic areas.	Limited availability (not widely available in Europe) [26]
HelioScope	A commercial web- and cloud-based software developed by Folsom Labs, primarily for designing PV installations. It offers basic power generation estimation based on location.	Paid software [27]
OpenSolar	A cloud-based solar design and sales platform that integrates with GIS data to streamline rooftop PV system assessments. It is heavily business-oriented and focuses on single-roof designs.	Free software [28]
Homer Energy	A desktop- and cloud-based platform simulating hybrid renewable energy systems, including solar, and evaluating financial feasibility based on weather and geodata. It focuses on off-grid and grid-connected projects.	Paid software (free trial available) [29]

Table 1. Overview of solar rooftop assessment tools.

Tool Name	Description	Availability
ArcGIS Solar Radiation toolset	These tools assess the suitability of rooftops for solar PV installations and incorporate GIS data layers such as building footprints, terrain models, and tree canopy data.	Paid software [23]
Photovoltaic Geographical Information System (PVGIS)	PVGIS is an online tool developed by the European Commission’s Joint Research Centre. It provides worldwide high-resolution solar radiation and PV performance data, allowing users to estimate energy yield of rooftop PV systems based on location-specific parameters.	Free software [24]
SOLARGIS	Developed by GeoModel Solar, the same researchers behind PVGIS, it offers extensive studies on solar irradiation, PV power estimation, and PV investment evaluation. It is a commercial company with numerous scientific outputs.	Paid software and services [25]
Project Sunroof	A web-based tool developed by Google using aerial imagery and 3D modeling to estimate solar potential of rooftops in specific geographic areas.	Limited availability (not widely available in Europe) [26]
HelioScope	A commercial web- and cloud-based software developed by Folsom Labs, primarily for designing PV installations. It offers basic power generation estimation based on location.	Paid software [27]
OpenSolar	A cloud-based solar design and sales platform that integrates with GIS data to streamline rooftop PV system assessments. It is heavily business-oriented and focuses on single-roof designs.	Free software [28]
Homer Energy	A desktop- and cloud-based platform simulating hybrid renewable energy systems, including solar, and evaluating financial feasibility based on weather and geodata. It focuses on off-grid and grid-connected projects.	Paid software (free trial available) [29]

Section 2.1	Software—A QGIS [30] application for data processing and other tools such as OpenStreetMaps [31] needed for the proposed methodology are described. The software choices are explained.
Section 2.2	Source Data—The specific data characteristics needed for this analysis and where to obtain the data.
Section 2.3	Solar Irradiation—The building shadow analysis using the r.sun [32,33] module is described, as well as how the final solar irradiation values for different rooftops are computed. This is the first key input of the proposed methodology.
Section 2.4	PV Power Output—The statistical analysis of a typical 1 kWp PV installation using PVGIS is presented. This is the second key input of the proposed methodology.
Section 2.5	Methodology—The complete diagram of the proposed methodology is introduced and described in detail.
Section 3	Results and Discussion—The results from the testing area in Bratislava, Slovakia, are presented using the proposed methodology. The section contains also a discussion about the methodology known limitations, while emphasizing its advantages.
Section Future Research	Future Research—Future outlook of the possible improvements of the proposed methodology are presented. These ideas are not definite and instead should be seen as a direction in which we see the future research to be going regarding our targets for open-source licensing and public data availability.
Section 4	Conclusions—The summary of the paper, highlighting the key advantages of the proposed methodology and presented results.

2. Materials and Methods

Proposed methodology is dependent on many input parameters, which are introduced in this section. Starting with software and source data and continuing with more sophisticated analyses on building shadows and PV power output. Finally, the methodology is introduced, and all its necessary parts are described.

2.1. Software

The main application chosen to process map files was QGIS [34], a free, open-source tool that includes all the necessary functions. It offers a wide range of available extensions, including the GRASS (Geographic Resource Analysis Support System) module [35], which provides advanced analysis capabilities that are crucial to determining the photovoltaic potential using our method, specifically using the r.sun function [36]. Data processing and analysis were performed using MS Excel and Python.

OSM was utilized to identify buildings and their categories on the raster map, specifically as a plugin in QGIS—“Quick OSM” [31]. For this application, INSPIRE Buildings database is a good alternative solution [37].

The PVGIS application [24] was used to determine the potential power output of installed photovoltaic plants.

Every software except MS Excel has an open-source license, not limited to any use. MS Excel was used as it is convenient to interpret the results, but it is not strictly necessary for the analysis itself.

2.2. Source Data

The primary data source was ZBGIS, a free web application that provides GIS data across the Slovak Republic [38]. The first ALS Project Cycle (2017–2023) was carried out over the entire territory of Slovakia, which was divided into 42 regions, where ALS (airborne laser scanning) was implemented progressively from west to east.

The scanning objectives included

• Data collection via aerial laser scanning,
• The processing and delivery of classified point clouds,
• The creation of the Digital Terrain Model (DTM) in raster format,
• The creation of the Digital Surface Model (DSM) in raster format,
• The quality control of point clouds and the DTM.

Mandatory quality criteria involved

• A scanning density of at least 5 points per m²,
• A minimum 20% overlap between scanning strips,
• The use of specified coordinate and height systems (S-JTSK, ETRS89),
• Absolute height accuracy of point clouds (≤0.15 m) and DTM (0.20 m).

The result of the laser scanning was a “point cloud” representing the Earth’s surface, which was used to create the DTM 5.0 by interpolating the “Ground” classified points. The output raster was provided in ASC, ESRI GRID, or TIFF format with a 1 × 1 m resolution. The DSM 1.0 model, generated from selected classifications, also had a 1 × 1 m resolution.

Independent quality control was conducted by both the supplier and the ÚGKK SR (Geodesy, Cartography and Cadastre Authority of the Slovak Republic) to verify compliance with specifications. Key control aspects included

• Scanning density,
• The overlap between scanning strips,
• Relative positional and height accuracy,
• Correct point classification into defined categories,
• The absolute positional and height accuracy of both point clouds and the DMR.

For DTM, structures taller than 0.30 m, vegetation, and underground objects were excluded. This model, along with the DSM (which included surface features like buildings and vegetation), provided a detailed digital representation of terrain and surface features across Slovakia.

The first and second ALS project cycles shown in

Table 2. Comparison of basic information and required parameters for the 1st and 2nd ALS cycles.

Parameter	1st ALS Cycle (DTM 5.0, DSM 1.0)	2nd ALS Cycle (DTM 6.0, DSM 2.0)
Cycle Duration	5 years	12 years
ALS Products	Classified point clouds, DTM, DSM	Classified point clouds, DTM, DSM
ALS Area per Location	900 to 1800 km2	500 to 900 km2
Total Locations	42	73
Scanning Period	Non-vegetation (Nov 1—Apr 30)	Non-vegetation (Oct 15—Apr 15)
Point Density (last return)	min. 5 points/m2	min. 15 points/m2
Beam Footprint	max. 0.25 m	max. 0.45 m
Overlap of Scanning Strips	min. 20%	min. 40%
Cross Strips	1 per flight	min. 3 per collection area
Point Cloud Classification	Mandatory—2 classes	Mandatory—10 classes
	Optional—10 classes	Optional—Wire, Tower
Vertical Accuracy of Point Clouds	$m_h$ ≤ 0.15 m	$m_h$ ≤ 0.10 m
Horizontal Accuracy of Point Clouds	$m_{XY}$ ≤ 0.30 m	$m_{XY}$ ≤ 0.20 m

, producing DTM 5.0 and DTM 6.0, respectively, differ in terms of cycle duration, participation, coverage, and accuracy standards. Notably, the second cycle improved spatial accuracy, increased point density, and expanded the required classification categories. The second cycle is currently in progress, so the first ALS project cycle data was used.

The DSM format was selected as the appropriate data format alternative as it includes raster data of the rooftop areas and is significantly smaller than points of clouds.

Regarding other public DSM datasets, multiple web portals offering public datasets are available. Some of them are mentioned in the

Table 3. Examples of publicly available DSM datasets.

Region	Data Description	Source
European Union	The official portal for European data, thousands of datasets.	[39]
United Kingdom	The LIDAR Composite DSM covering  99% of England at 1 m spatial resolution.	[40]
Scotland	Various high-quality DTM, DSM and DEM datasets produced from point cloud data.	[41]
Luxembourg	This DSM is the result of a first LIDAR survey flight that has been done in October 2017.	[42]
Washington, DC (USA)	On this site, the District of Columbia government shares hundreds of datasets.	[43]
Vienna (AUT)	A very detailed mapping of the Austrian capital city, various datasets.	[44]

below.

2.3. Solar Irradiation

The intensity of solar radiation on the roof surfaces and the shaded parts of the buildings, unsuitable for installation, was obtained using the r.sun function of the GRASS module. This function considers roof parts or entire roofs that other objects will shade during solar energy exposure. It determines the intensity of incoming solar radiation based on input maps of object orientations relative to cardinal directions and their slopes. This analysis is the first key input of the proposed methodology.

This analysis was carried out in each season in the year, namely in February, May, August, and November. As expected, there are huge differences in the building shadows over the year, as well as the maximum irradiation over the examined urban area.

Through this analysis, during calculations, separate areas/pixels can be categorized into more and less efficient. Since the resolution of a single pixel is exactly 1 m², it makes it easy to calculate the rooftop areas with the most perspective. This allows the adjustment of the electricity production deviations from the reference values of southern roofs receiving the highest solar radiation, which is described later in this paper.

Figure 2 and Figure 3 show the results of the r.sun function from the GRASS module. The lighter the color, the higher the irradiation value of the pixel. These data are then filtered based on the roof placement and categories taken from the OSM. The histograms show a clear distinction between two main seasons—“summer” and “winter”. During the “winter” season, the most common values from global horizontal irradiance are around 400–500 Wm⁻², while during the “summer” season, it is around 900 Wm⁻², and the global maximum is higher. These values are snapshots during the middle of the examined day, but they were later extrapolated throughout the year with the help of the PVGIS tool. The high spikes in historgrams around the value 100 Wm⁻² are the generally the lower areas in the shadow, near higher structures like building, trees, or other objects. These areas are mostly filtered by the rooftop placement—the rooftop area is only 8.11% of the total examined area. However, some roofs are still heavily shadowed during the “winter” season, which is discussed later in this paper.

2.4. PV Power Output

The potential energy output from rooftop photovoltaic systems was assessed using the PVGIS API service (Figure 4) [24], providing calculations for the total power output under ideal conditions in terms of the smooth operation of the PV installation without any technical issues. Production deviations will be determined in the final calculations based on the output of the r.sun function mentioned earlier and the production coefficients of the photovoltaic system according to its orientation [45]. Optimized angles and orientations for rooftop installations were considered as the baseline data for power output calculations.

A statistical analysis was carried out over period of 16 years, from 2005 to 2020, when the average PV production and power output were estimated. During the years, it can be observed that the “summer” season could be defined as starting in March/April end lasting until August/September, while the rest of the year is considered the “winter” season. These findings perfectly align with the r.sun model results.

As can be seen in more detail (Figure 5), individual years can vary a lot, compared to the statistical results in Figure 4. In this specific year, the most promising month for PV production appeared to be in April, but this is not correct, as is generally said. Every month from April to August could potentially be the one with highest irradiation values, depending highly on the weather that year. More rainy and cloudy summer would result in higher potential power outputs in spring/autumn, and vice versa.

Another observation can be made—the green dashed lines represent different non-ideal vertical alignment of the PV panel. Different orientations behave differently under these circumstances. Also, the basic verification of the output data is clearly visible—the west- and east-oriented rooftop installations have an opposite character of this vertical alignment effect on power output. This is logical because, throughout the day, the sun is moving along the east–west axis. Also, the north orientation suffers the most, when the non-ideal slope of the roof is analyzed (compared, for example, to the south orientation, where the slope of the roof has minimal impact).

The effect of sloped roof in terms of different aspects is summarized in

Table 4. PV panels efficiency based on their orientation and slope.

Orientation/Slope [°]	20	30	40	50	60
North	64%	54%	44%	36%	29%
East	81%	79%	77%	73%	68%
South	96%	99%	100%	98%	95%
West	81%	79%	76%	73%	68%

and is directly tied to the results of the analysis performed in Figure 5. Similar tabulation can be found for several generalized areas; however, it is always advised to use the most case-specific data available, especially when taking into the consideration the character of the solar irradiance data. This effect of sloped roofs is reflected in Figure 5 by the green dahsed lines—where the non-ideal slopes are affecting the resulting PV output power. Another validation of the analysis is the general characteristics of the daytime sub-graphs on the left side of Figure 5—the South orientation clearly has the most efficient orientation, the North has the worst efficient orientation, and the West and the East mirror themselves in their shape of load curves.

This analysis emphasizes the need for proper input data validation and proper data preparation. This statistical analysis is the first key input of the proposed methodology, because it reflects the expected behavior of the average 1 kWp system in the area. The efficiency of the PV system is already included in the PVGIS results. The partial results from Table 4 validate the expected efficiency loss during the sub-optimal orientation or slope of the system. However, this table is not used further in the computation, since much more precise irradiance values are taken from the r.sun analysis, which is described in the previous section.

2.5. Methodology

A general flow diagram of the proposed methodology, which utilizes the previously mentioned software and data sources, is illustrated in Figure 6. The primary concept of this methodology is its broad applicability and relative ease of use, coupled with precise estimates of photovoltaic system power output. In contrast to similar research that typically uses only reduction coefficients for estimating photovoltaic power, as referenced in [17,45], this paper introduces an analysis of building shadows.

The first task is to import the map data into the QGIS project [34]. After ensuring the correct coordinate systems of the imported data, it is possible to utilize the slope and aspect analyses. The QGIS GDAL (Geospatial Data Abstraction Library) library and the GRASS GIS module offer efficient methods for these analyses. The r.slope.aspect function in the GRASS module can perform these analyses simultaneously.

For the r.sun [46] function (building shadow analysis) in the GRASS module, a separate project must be created with a “GRASS mapset” file to specify the coordinate system used for GRASS analyses. Input data for the r.sun function include a base raster map in DSM format and two layers for slope and aspect analysis. In addition, it is necessary to specify the date and time for the analysis. This is a simplification of the article, where only one hour was used in the proposed methodology. In further parts of the paper, other alternatives to building shadow analysis are presented.

The OSM vector map of the area of interest represents all objects as points, lines, or individual/combined polygons. This layer contains a large amount of information assigned to each object in the attribute table, detailed building identification numbers, and their usage types. Using the “Quick OSM” plugin in QGIS, it is possible to highlight buildings from the OSM vector map.

The final steps involve processing raster layers of solar radiation intensity and classified slopes. Combining attribute tables from three layers into one and exporting it to CSV format is carried out using the “Join attributes by location” command. This command can combine data based on the (x,y) coordinates of each polygon and may need to be executed multiple times.

The OSM building types, classified as shown in

Table 5. Roof pitch categories.

Slope Category	Slope Angle [deg]	Description
1	0–10	flat roof area
2	10–45	lightly slanted roof area
3	45–60	heavily slanted roof area
4	60+	removed roof area

and

Table 6. Building types according to OSM and their categorization.

Building Type (OSM)	Commercial	Residential	Removed
yes	-	-	×
apartments	-	×	-
detached	-	-	×
service	×	-	-
garages	×	-	-
roof	×	-	-
school	×	-	-
civic	×	-	-
commercial	×	-	-
government	×	-	-
residential	-	×	-
office	×	-	-
garage	×	-	-
university	×	-	-
hospital	×	-	-
retail	×	-	-
industrial	×	-	-
house	-	×	-
carport	×	-	-
semidetached_house	-	×	-
kindergarten	×	-	-
train_station	×	-	-
construction	-	-	×
parking	-	-	×
hotel	×	-	-
college	×	-	-
clinic	×	-	-
sports_centre	×	-	-
silo	-	-	×
church	-	-	×
dormitory	×	-	-

and Figure 7, are directly inspired by statistics from [45]. As can be seen in Figure 7, most buildings were categorized as apartments (orange—secondary axis), followed by retail and offices (blue—primary axis). It is evident that more than a third of all objects in the selected area are removed, the largest being unrecognized objects, “yes”, which represents 96% of the removed areas. Therefore, by obtaining data that would provide a more accurate classification of buildings, the results of the installed power potential in the selected area can be significantly improved. Other roofs that are considered removed are those with too little total solar irradiance. There can be multiple reasons for these low irradiance values, such as the orientation of the roof to the north while having enough steepness or being in the shadow of another higher building nearby. In all these cases, the r.sun functionality is to evaluate the irradiance values properly.

Photovoltaic systems power estimation is carried out as follows (regarding to

Table 7. Settings and parameters of the model.

Model Settings and Parameters
$c_{area\_flat}[-]$	1.0
$c_{area\_<45}[-]$	1.2
$c_{area\_<60}[-]$	1.7
$c_{occ\_flat}[-]$	0.67
$c_{occ\_slanted}[-]$	0.80
$A_{PVS\_1\mathrm{kWp}}$ [m2]	4.600
$E_{PVS\_1\mathrm{kWp}\_yearly}$ [MWh]	1.179
$GH{I}_{high}$ [Wm−2]	600
$GH{I}_{low}$ [Wm−2]	300
$GH{I}_{max}$ [Wm−2]	1083
$c_{GHI\_high}$ [−]	0.78
$c_{GHI\_mid}$ [−]	0.42

).

• Definition of the minimum area of the 1 kWp photovoltaic installation: This area was estimated at $A_{PVS\_1\mathrm{kWp}}$ = 4.6 m². The most common sizes of solar panels Longi Solar LR5-72HIH were estimated.
• Definitions of the slanted roofs area correction coefficient are introduced with similar methodology as in [45,47]; they are shown in Table 7. The general correction formula for the roof area is as follows:

  $$A_{PVS\_full}=c_{area\_flat/<45/<60}·A_{PVS\_area\_flat/<45/<60}\left[{\mathrm{m}}^2\right]$$

  (1)
• Definition of the rooftop occupancy coefficients $c_{occ\_flat/slanted}$: According to multiple sources and by brief verification in the respective dataset, the coefficients were $c_{occ\_flat}=0.67$ and $c_{occ\_slanted}=0.8$. To evaluate this in more detail, a deeper analysis should take place considering also the type of building and most likely utilizing a more complex automation process than the workflow presented in this paper. A final value of $A_{PVS\_real}$ [m²] is achieved.

  $$A_{PVS\_real}=c_{occ\_flat/slanted}·A_{PVS\_full}\left[{\mathrm{m}}^2\right]$$

  (2)
• The photovoltaic power estimation is calculated from the individual slope areas of the analyzes performed. The sum of installed photovoltaic power is the sum of all areas.
• The yearly energy production of the generic photovoltaic system is obtained from the PVGIS application (Figure 4), corrected by the r.sun building shadow analysis. The maximum photovoltaic potential is therefore on the roofs where there are no shadows and are oriented to the optimal direction (South). The PVGIS application computes the power produced by the modeled photovoltaic system throughout the year, given the changes in the total irradiance in the area. This is used as a baseline enhanced with the r.sun shadow analysis.
  As can be seen from Figure 8, the separation of the GHI parameter to multiple windows enables a better evaluation of the PV potential over the urban area. The graphs show the composition of this parameter during each of the examined months in both seasons. The middle range of the GHI is the most equally presented in both the residential and commercial building category. These graphs in general confirm the resulting histograms from the r.sun function, which can be seen in Figure 2 and Figure 3. It is clearly visible how few areas are sufficiently irradiated during the winter season.
• The power output and energy production are scaled according to the average global horizontal solar irradiance from the seasonal case studies. The maximum observed GHI was 1083 Wm⁻², and limits for “high” and “low” solar irradiance values are stated. Using these values, the correcting coefficients for the power output of different GHI categories are introduced—$c_{GHI\_high}$ for GHI values over 600 Wm⁻² and $c_{GHI\_mid}$ for GHI values between 300 and 600 Wm⁻². The areas with lower GHI values then 300 Wm⁻² were removed from any power output estimations. As can be seen in Figure 2 and Figure 3, there are large differences between the shadows of the building that are cast throughout the year.

  $$P_{PVS\_real}={\displaystyle \frac{A_{PVS\_real}}{A_{PVS\_1\mathrm{kWp}}}}\left[\mathrm{kWp}\right]$$

  (3)

  $$E_{PVS\_real}=P_{PVS\_real}·E_{PVS\_1\mathrm{kWp}\_yearly}·c_{GHI\_high/mid}\left[\mathrm{MWh}\right]$$

  (4)

  $$P_{PVS\_sum}=\sum _{i=1}^n{P}_{PVS\_real\_i}\left[\mathrm{kWp}\right]$$

  (5)

  $$E_{PVS\_sum}=\sum _{i=1}^n{E}_{PVS\_real\_i}\left[\mathrm{MWh}\right]$$

  (6)

3. Results and Discussion

This paper analyzed four case-study scenarios, each in different months equally distributed over the whole year. To determine all shaded areas perfectly, even this analysis is insufficient. Shaded areas change throughout the day depending on the angle and position of the sun, significantly increasing during the winter season because the sun is low in the sky and the direction of solar radiation is much lower than in summer. A more detailed comparison can be seen in Figure 9. The property of the r.sun routine is that it can only perform analyses during a specified day and hour, so it is not possible to simulate all roof areas that will be shaded throughout the year in one simple execution. The solution to this problem relies on combining many analyses during different days and hours throughout the year. The result of the r.sun analysis is a raster layer containing solar radiation intensity values [Wm⁻²], from the minimum values in completely shaded areas in black to the maximum values in white. However, the results of this paper are based on the statistical evaluation of data, which is a substitute for this complex methodology. Therefore, the hardware (and time) requirements are kept relatively low, while keeping the results in realistic numbers.

The complete time needed to run every mentioned analysis over one raster datapoint on an average laptop (Intel i5-1135G7 2.42 GHz, 16 GB RAM, Windows 11-64 bit) is circa 60 min, while the most time is required for the combination of attribute tables (circa 80%). However, this time can be improved by automating the process and efficient script writing. This is planned as one of the future activities of this research group. Nonetheless, it was demonstrated that low hardware requirements are needed for a relatively detailed analysis over significant urban area while using the open-source applications and data.

As an example of the slope and aspect analysis, two pictures are shown in Figure 10. They illustrate how the categorization of these parameters significantly aids in filtering minor discrepancies in the source data—every identified area has every parameter in the final attribute table. However, the most impacting factors in the conducted analyses were still the resulting irradiation data from the r.sun analysis. The final created attribute table for each rooftop area is the core of this methodology.

The results of the photovoltaic potential obtained by this method can be divided into installed photovoltaic power and total annual production of photovoltaic energy—summarized in the tables below.

Table 8. Results—residential roof area [m²].

GHI	Summary (m2)	Flat (m2)	<45° (m2)	<60° (m2)	>60° (m2)
HIGH	84,309.49	21,940.45	52,192.75	6155.60	4020.69
MID	22,388.05	10,925.52	8172.36	2593.44	696.73
LOW	96,514.82	14,646.74	57,480.65	12,049.37	12,338.07
sum	203,212.36

,

Table 9. Results—commercial roof area [m²].

GHI	Summary (m2)	Flat (m2)	<45° (m2)	<60° (m2)	>60° (m2)
HIGH	42,762.51	21,714.60	16,546.50	2386.71	2114.70
MID	10,746.31	6211.20	2847.16	1101.73	586.23
LOW	59,907.89	25,976.34	16,581.73	5120.40	12,229.42
sum	113,416.72

and

Table 10. Results—residential + commercial roof area [m²].

GHI	Summary (m2)	Flat (m2)	<45° (m2)	<60° (m2)	>60° (m2)
HIGH	127,072.00	43,655.05	68,739.25	8542.31	6135.39
MID	33,134.36	17,136.72	11,019.51	3695.17	1282.96
LOW	156,422.71	40,623.07	74,062.38	17,169.76	24,567.49
sum	316,629.08

contain the suitable roof areas for each building category, while

Table 11. Results—residential power output.

$\mathit{GHI}$	${\mathit{P}}_{\mathit{PVS}\_\mathit{sum}}$	${\mathit{E}}_{\mathit{year}\_\mathit{sum}}$	${\mathit{P}}_{\mathit{PVS}\_\mathit{flat}}$	${\mathit{E}}_{\mathit{year}\_\mathit{flat}}$	${\mathit{P}}_{\mathit{PVS}\_<45}$	${\mathit{E}}_{\mathit{year}\_<45}$	${\mathit{P}}_{\mathit{PVS}\_<60}$	${\mathit{E}}_{\mathit{year}\_<60}$
[-]	$\left[\mathbf{MW}\right]$	$\left[\mathbf{GWh}\right]$	$\left[\mathbf{kW}\right]$	$\left[\mathbf{MWh}\right]$	$\left[\mathbf{kW}\right]$	$\left[\mathbf{MWh}\right]$	$\left[\mathbf{kW}\right]$	$\left[\mathbf{MWh}\right]$
HIGH	19.97	16.56	3195.67	2927.53	10,892.40	9978.45	1819.92	1667.21
MID			1591.33	779.57	1705.54	835.52	766.76	375.63
sum			4787.00	3707.11	12,597.94	10,813.97	2586.67	2042.84

,

Table 12. Results—commercial power output.

$\mathit{GHI}$	${\mathit{P}}_{\mathit{PVS}\_\mathit{sum}}$	${\mathit{E}}_{\mathit{year}\_\mathit{sum}}$	${\mathit{P}}_{\mathit{PVS}\_\mathit{flat}}$	${\mathit{E}}_{\mathit{year}\_\mathit{flat}}$	${\mathit{P}}_{\mathit{PVS}\_<45}$	${\mathit{E}}_{\mathit{year}\_<45}$	${\mathit{P}}_{\mathit{PVS}\_<60}$	${\mathit{E}}_{\mathit{year}\_<60}$
[-]	$\left[\mathbf{MW}\right]$	$\left[\mathbf{GWh}\right]$	$\left[\mathbf{kW}\right]$	$\left[\mathbf{MWh}\right]$	$\left[\mathbf{kW}\right]$	$\left[\mathbf{MWh}\right]$	$\left[\mathbf{kW}\right]$	$\left[\mathbf{MWh}\right]$
HIGH	9.15	7.60	3162.78	2897.40	3453.18	3163.44	705.64	646.43
MID			904.67	443.19	594.19	291.09	325.73	159.57
sum			4067.45	3340.59	4047.37	3454.52	1031.37	806.00

and

Table 13. Results—residental + commercial power output.

$\mathit{GHI}$	${\mathit{P}}_{\mathit{PVS}\_\mathit{sum}}$	${\mathit{E}}_{\mathit{year}\_\mathit{sum}}$	${\mathit{P}}_{\mathit{PVS}\_\mathit{flat}}$	${\mathit{E}}_{\mathit{year}\_\mathit{flat}}$	${\mathit{P}}_{\mathit{PVS}\_<45}$	${\mathit{E}}_{\mathit{year}\_<45}$	${\mathit{P}}_{\mathit{PVS}\_<60}$	${\mathit{E}}_{\mathit{year}\_<60}$
[-]	$\left[\mathbf{MW}\right]$	$\left[\mathbf{GWh}\right]$	$\left[\mathbf{kW}\right]$	$\left[\mathbf{MWh}\right]$	$\left[\mathbf{kW}\right]$	$\left[\mathbf{MWh}\right]$	$\left[\mathbf{kW}\right]$	$\left[\mathbf{MWh}\right]$
HIGH	29.12	24.17	6358.45	5824.93	14,345.58	13,141.88	2525.55	2313.64
MID			2496.00	1222.76	2299.72	1126.61	1092.49	535.20
sum			8854.45	7047.70	16,645.31	14,268.49	3618.04	2848.84

provide complete info about the potential PV power output and annual energy yield from the rooftop PV installations. All these results reflect the average values obtained from all the case studies conducted in this analysis. Taking every month separately would result in too many data without little additional information value.

The summarized results are shown from a different perspective in Figure 11, Figure 12 and Figure 13. The changes in different observed seasons are clearly showing the expected character of the PV power output throughout the year.

Future Research

There are multiple options for future research, but the main one is greater automation in QGIS data processing. This would allow for more complex studies in terms of longer time periods and even larger urban areas. In addition, these studies can also be performed from point cloud data [48,49], which should enhance the precision but extend the computational time considerably. The results could be directly related to the power consumption of the urban area [50,51], which should improve the general estimation of the PV impact in the urban area; however, it is questionable how to properly estimate the selected area’s consumption, and if the area is too wide, it could even be impossible to include. The general increase in electricity prices [52] could form the baseline for various economic analyses over the area. The modeling of photovoltaic systems with more precision, such as in [53,54], would replace the PVGIS application in the process but should first be properly verified. Another option could be an extended shadow analysis algorithm [55], which considers the shadows of the PV panels on the flat rooftops or even the more precise solar irradiation methodology investigation [56], which will enhance the r.sun module results. These new possible improvements may, however, result in more limited usage of the methodology, mostly in hardware requirements, or even in the open-source licensing of the data or software used. As these parameters were the initial expectations of the proposed methodology, there will be an ongoing effort to also stick to them in the future research.

4. Conclusions

This paper presents a direct and open-source approach to estimate the annual energy output and power generation of rooftop photovoltaic systems. The study was carried out to showcase its efficiency and results in a representative urban sector of Bratislava. Relevant data sources were pinpointed, and a detailed procedure of the required analyzes was provided. The mentioned constraints are tied to statistical and practical aspects, but they leave room for potential improvements, encompassing one of the most significant and common issues such as sustaining models and analyzes that are efficient, scalable, and easy to use.

The advantage of the proposed methodology iss definitely its straightforward application in practice. Both key inputs (PVGIS statistical analysis and the solar irradiation r.sun analysis) are the baselines for any PV study over the selected area. The presented results shows the estimation of the maximum PV power output considering the building shadow analysis and realistic assumptions about the rooftop occupancy. Each analysis is described and explained, so as to be replicable on any publicly available dataset.