An approach to the design of photovoltaic noise barriers and a case study from Istanbul, Turkey

Solar energy solutions that do not require additional space are critical. Noise barriers, which are built in low-value lands next to noise sources, provide effective areas for PV modules. There are many studies on using noise barriers as a sub-structure for photovoltaic systems, providing electricity generation besides noise reduction targets. Photovoltaic noise barrier (PVNB) technology combines noise control measures with renewable energy generation. In this study, it is aimed to develop an integrated design method that embeds solar energy technology in noise protection structures. The method is exemplified in an existing settlement located on the side of the road with heavy traffic. According to local climate and solar data, optimum tilt angles have been determined for annual, semi-annual, seasonal, and monthly periods. Noise barrier alternatives are derived with combinations of different diffraction edge sizes of barrier top and determined optimum inclination angles. The performance of the criteria that affect the PVNB effectiveness for alternatives was calculated through software tools. The energy generation potential of PVNB and its shading in adjacent blocks were calculated with PVsyst 6.7.7. The noise control efficiency of the structure was computed via SoundPLAN 7.2. TOPSIS method, one of the most common multi-criteria decision-making (MCDM) techniques, was used in the evaluation. As a result of TOPSIS, the best PVNB solution in the case study is the alternative that has 3-m and 2-m-wide edges and 58° and 31° tilted edges. In comparison with the current situation, the selected alternative will decrease 44% of the number of receiving points affected by noise and provide 524,804 kWh annual electricity generation.


Introduction
The significant positive impact on carbon dioxide emissions has been supported by empirical studies that shape the energy matrix of the world with renewable energy sources (Acheampong et al. 2019;Al-Mulali et al. 2015;Apergis et al. 2010;Inglesi-Lotz and Dogan 2018;Khoshnevis Yazdi and Ghorchi Beygi 2018). Among the renewable resources encouraged by greenhouse gas reduction policies, solar energy has a resource potential that technically exceeds all global energy demand (Ueda et al. 2008).
Power density is the maximum amount of power that can be generated in a land area. Solar energy is low efficient relative to fossil fuels, due to its low energy density (Topcu and Ulengin 2004). Therefore, the need for large areas of land in the supply of solar energy creates a disadvantage for this technology (Calvert and Mabee 2015;Dijkman and Benders 2010;Graebig et al. 2010;Nonhebel 2005;Rathmann et al. 2010). Solar energy solutions that do not require more land are important when considering that world population growth puts pressure on food and energy. Examples of these solutions are hybrid systems where photovoltaic systems are applied with different renewable energy technologies on mutual land (Calvert and Mabee 2015;Li et al. integrated photovoltaic systems (Biyik et al. 2017;Jelle et al. 2012;Peng et al. 2011), agrovoltaic systems (Dupraz et al. 2011), and PV systems on low-value lands like brownfields (Denholm and Margolis 2008). Besides, photovoltaic system solutions that do not require additional space by integrating noise barriers ensure the double use of land resources (Nordmann and Clavadetscher 2004).
Noise barriers can be the best solution when no noise control measures are taken neither at the noise source nor at the receiver (Garg et al. 2013). Noise barriers are critical in providing acoustic comfort between major thoroughfares and settlements along roads. Physical environmental factors, which have negative effects on human health and comfort, can cause serious disturbances in urban areas (Amoatey et al. 2020;Zorer Gedik et al. 2017). Paschalidou et al. emphasized that mitigation measures are especially important during the night because excessive night-time exposure to noise has the biggest harmful effect on public health (Paschalidou et al. 2019).
In the presence of a barrier, noise from a source reaches the receiver through two paths: diffracted waves at the top edge of a barrier and the transmitted pathway through the barrier. Transmitted sound is negligible as barriers are constructed of solid materials. As a result, the barrier performance is limited by the diffracted sound (Fard et al. 2013). Many studies have been conducted on the height and width of the noise barriers, since the diffraction path between noise source and receiver is related to height (Maekawa 1968). Calculation of Road Traffic Noise (CRTN), the method used in predicting road traffic noise, assumes that a barrier has an insignificant thickness, but diffraction over the top edge of a barrier is affected by its cross-section. Providing a substantial reduction in noise level in the shadow zone without increasing the height of the barrier is the focus of the studies. It is impractical to build barriers of extensive height due to aesthetic, economic, and safety reasons (Grubeša et al. 2012). The issue of designing multiple edges on a barrier top has been the focus of many researches. It has been observed that adding profiles to the top edge of the vertical barrier increases the insertion loss without a higher noise barrier and absorber surfaces (Crombie and Hothersall 1994;Watts 1996). Many authors researched for the determination of noise barrier tops such as Y, T, arrow, and cylinder, using experimental and numerical methods (Baulac et al. 2008;Greiner et al. 2010;Hothersall et al. 1991a, b;Hothersall et al. 1991a, b;Monazzam and Lam 2008). T-shaped barrier designs are effective in noise reduction as the diffraction surface is close to the noise source (Hothersall et al. 1991a, b;May and Osman 1980). It was also determined that when the slope of the arms of the T-shaped barriers is 120°, it is the most effective solution in reducing low-frequency sound (Venckus et al. 2012). Ho et al. and Shao et al. have determined that noise barriers with random edge profiles offer more insertion loss than flat ones, especially at high frequencies (Ho et al. 1997;Shao et al. 2001). Crombie et al. found a better attenuation for barriers with two or more diffraction edges by numeric methods (Crombie et al. 1995). Another examination has shown that when multiple diffraction edges are used, the surfaces are effective when they are rigid rather than absorbent (Ishizuka and Fujiwara 2004).
In hybrid solutions, where solar energy and noise control are handled together, photovoltaic panels are integrated into the noise barrier with an inclination angle to get maximum power. The inclination of the PV arrays should be determined according to data such as climate data, shading condition, energy consumption profile, structural constraints, and acoustic performance of barrier. The amount of solar radiation reaching the solar modules affects the energy generated in the photovoltaic systems (Huld et al. 2011). The radiation intensity of the module is predominantly determined by the panel orientation and panel tilt. The panel inclination should be deliberated at the design stage, since the orientation of the road that determines the azimuth angle of PVNB systems is fixed and unchangeable. The optimal tilt angle of a solar collector is related to the local climatic condition, the geographic latitude, and the period of its use (Bakirci 2012). On the other hand, Wadhawan and Pearce highlight that non-optimal configurations also offer the potential for large fractions of electricity generation, when soiling losses and structural costs of PVNB systems are analyzed together with other system parameters (Wadhawan and Pearce 2017). When the annual energy output data of the PVNB installations in Germany are examined, it has been observed that near-vertical inclined panels with airflow and less shading loss can perform better than panels with optimum tilt (Nordmann and Clavadetscher 2004). Both approaches were used in the PVNB design, on the side of the A22 Brennero motorway in Italy. It comprises two edges with different tilts: one of the diffraction edges of the structure has an optimum angle of inclination (35°), while the other has a steeper angle (60°) (Costa and Duiella 2010). The tilt angles should also be chosen high enough to hinder soil accumulation. Jaszczur et al. stated that dust accumulation on panel surfaces is a complex phenomenon that depends on a large number of different environmental and technical factors. Also, in the experimental study conducted in urban area on panels with 15° and 35° inclination, the maximum dust density was recorded for panels with 15° inclination (Jaszczur et al. 2020). Lu and Zhao have observed the maximum dust deposition rates for the tilted PV panel angles of 25°, 40°, 140°, and 155°, respectively (Lu and Zhao 2018). Vallati et al. concluded that when the average height of the roadside buildings is low (3-6 m), T-shaped barriers will be best. When the average height of the building is higher (6-15 m), it has obtained the best result in terms of noise control and solar energy performance in PVNB with a diffraction edge inclined to 60° (Vallati et al. 2015).
Environmental impact assessments (EIA) play an important role in newly planned public projects, as well as the technical performance of the design (Tamura et al. 1994). A sound barrier, as part of the surrounding landscape, could be a cause of impact for both drivers and residents along the barriers. Noise barriers also create some disadvantages, in addition to providing acoustic comfort for residents living behind obstacles. For example, there is a study which has negative views of residents such as landscape obstruction, a feeling of being trapped, loss of air circulation, loss of sunlight, and inadequate maintenance of the barrier (Tamura et al. 1994).
In this study, a novel design method in which it integrates solar energy technology into noise barrier solutions has been developed. The method was exemplified in Istanbul. Within the study, an existing residential settlement exposed to the noise level above the limit value, along a highway with heavy traffic, was considered. The flow chart showing the steps of the case study is shown in Fig. 1. Noise maps of the current environment were generated by using SoundPLAN 7.2 simulation software, and the accuracy of the maps was validated by comparing them with actual noise measurements. Barrier design and design variables were determined with literature review and site-specific needs. In compliance with local climate and solar data, monthly, seasonal, semi-annual, and annual optimum tilt angles have been determined. Combinations of determined optimum tilt angles with diffraction edges in different sizes have been grouped as alternatives. The energy generation potential of barrier alternatives and light loss in neighboring blocks were calculated with PVsyst 6.7.7. The acoustic performance of the noise barriers was tested with the SoundPLAN 7.2 software. The best alternative for the site was determined by analyzing the barrier alternatives with TOPSIS, which is a multi-criteria decision method.

Methodology
The following steps outline the case study: -Gathering data and making assumptions.

Gathering data and making assumptions
The study is based on the assumptions shown in Table 1 and listed below: -As stated in the literature review, multiple edge barriers are effective when the surfaces are rigid. Due to the rigidity of the photovoltaic panel surfaces, the barrier type with 2 diffraction edges was determined as the optimum form.  -To show the effectiveness of the PVNB alternatives, some parameters have been kept constant, including traffic characteristics, atmospheric conditions, road-barrier distance, and topographic conditions. -It is assumed that the PVNB will be built uninterruptedly between two secondary roads included in the actual maps of the study field. -It is envisaged that the PVNB installation will be used as an electric vehicle charging station for residents. Due to the location of the PVNB in the area designed for a parking lot, the barrier top has been 5 m wide in line with parking lot length. -At the analysis step of PVNB alternatives in terms of noise control, the situation in which the PVNB is positioned instead of the garden wall is accepted as the basic case to make the assessment more apparent. -While determining the number of people affected by noise, it is accepted that the apartments are used at full capacity and there are no vacant flats. -The PVNB barriers were designed as the lower vertical part (height equal to 1 m from the road surface) made out of precast concrete panels on which uprights are installed. The mass of the lower part of the barrier should be a minimum of 20 kg/m 2 or more for a transmission loss of at least 20 dB(A) for adequate sound transmission loss (FHWA 2000). -The lower part of the barrier facing the roadside was designed with sound-absorbing panels (load-bearing structure + sound-absorbing panels) to minimize reflections.
The methodology was validated in two stages for solar energy and noise control. First, solar irradiation outputs were validated by comparing literature review outcomes with data concerning. Upon completion of the literature review on the forecast performance of the software used in the performance analysis of solar power plants at different locations with variable scales, it was concluded that PVsyst is the most realistic and effective software (Mermoud 2011;Silva et al. 2020;Umar et al. 2018). Second, acoustic environment modeling was validated by noise measurements at the case study site.

Modeling
The noise maps, showing the acoustic climate of the 450,000 m 2 area where the case study was carried out, were generated in SoundPLAN 7.2 (Braunstein + Berndt GmbH) software (SoundPLAN n.d.). The base maps of the site and the environmental properties were transferred to SoundPLAN 7.2. Population data and properties of noise sources were assigned in the software. The number of residents was calculated according to the room numbers of the flats and defined in the software as given in Table 2. The distances of the blocks in the first row most affected by the noise from the highway vary between 20 and 70 m. Figure 2 appears the site plan of the case study plot.
The 1:1000 scale maps of the site were provided by the Municipality of Çekmeköy in the digital environment. SoundPLAN calculated the digital ground model (DGM) from the elevation data, after importing the actual map with terrain points for each coordinate in dxf format. The number of floors was modeled using municipal maps as the basis and story height was assigned as 3 m in the software in line with the field data.
Annual data on the traffic volume of the highway was got from the Istanbul Department of Transportation. Under data collected from the closest sensor (sensor no: 551), the total number of vehicles passing the highway between 31.05.2018 and 31.05.2019 is 34,946,809. Traffic flow characteristics of the highway are given in Table 3 according to the daytime, evening, and nighttime intervals. The traffic volume on the highway does not differ significantly on a monthly and daily basis. Daytime vehicle volume converted to annual average daily traffic (ADDT) and a daily traffic volume of 95,744 vehicles/day. The road is one of the highways with the highest density in the traffic volume scale of the General Directorate of Highways. Traffic data of other secondary roads in the district were also taken into account in noise mapping.
The climate data of Istanbul were defined in SoundPLAN 7.2 software. Prevailing and secondary wind direction was defined to be NNE and SSW, respectively; the mean annual temperature was defined to be 13.9° and average relative humidity was defined to be 71.5% (Climate consultant 6.0 n.d.).
The dual carriageway is defined into software as 3.5-m lane width and 0% inclination. Smooth asphalt pavement was selected as the road surface in the software. In pursuance of the European Union Environmental Noise Directive, noise maps should present the sound level distribution at 4 m above ground (European Parliament and Council of the European Union 2002). However, the ANNEX 1 of the same directive states that other heights may be chosen for design local measures, provided that the minimum height above ground should be 1.5 m. Since this study evaluates the acoustic attenuation in open areas, assessment point height was accepted as 1.5 m, the average height of the ear level (standing) (Akdağ et al. 2017).
The grid spacing in noise mapping should be no more than 10 m, even 5 m spacing may be desirable in urban areas (WG-AEN 2007). In this study, the grid space was selected as 5 × 5 m as stated in the guide since the case study was conducted in an urban zone. The daytime noise indicator Lday time interval (07:00-19:00) was taken into consideration to evaluate the noise mitigation simultaneously with the solar potential of PVNB alternatives. Since solar panels do not produce energy in the evening and at night, the evening (Le-19:00-23:00) and night (Ln-23:00-07:00) noise maps have not been generated.

Validation
In the case study, actual noise measurements and acoustic simulations of the existing area were performed. Grid noise map, facade noise map, and cross-section map were   Figure 3 reveals the current noise climate of the case study area. Traffic noise modeling and traffic noise measurements are complex tasks, due to the stochastic features of traffic. The measured noise level was got by adding the noise generated by independent noise sources (industry, inhabitants, animals, etc.) to the traffic noise (Prezelj and Murovec 2017). Therefore, the validation step is critical to ensure that the noise model is an accurate representation of the real acoustic environment. The most appropriate way of the validation process is based on the comparison between actual noise measurements and simulated model outputs with their respective point receivers (İlgürel et al. 2016). The difference between the measurement data and simulation outputs and the standard deviation values of the data series is called uncertainty indicators.
The validation process that takes into account the uncertainty of a noise measurement ensures the most accurate representation of the model (Murillo Gómez et al. 2020). The uncertainty range or the validation threshold, defined as a quantity of reliability of the calculated values, has been specified in some experimental studies. In their study, Maruyama et al. observed that, if the number of vehicles passing during the measurement time interval exceeds 170, uncertainty indicator (ΔLAeqT) must be within the range of ± 1 dBA (Maruyama et al. 2013). In some studies, when the measured and calculated data are compared, depending on the noise estimation model used in the study, the validation threshold should be between ± 3 and ± 5 dBA (Bastián-Monarca et al. 2016;Lee et al. 2008;Vukadin et al. 2008). According to Good Practice Guide for Strategic Noise Mapping and the Production of Associated Data on Noise Exposure, the difference between actual and calculated results should not exceed 1 dBA in a distance of 300 m from the source, 3 dBA in a distance of 600 m from the source, and 10 dBA in a distance of 2,000-3,000 m from the source (WG-AEN 2007).
The noise level measurements made in the case study area were performed as per ISO 1996-1 standard (ISO 1996(ISO -1:2003. Acoustics -Description, of environmental noise -part 1: basic quantities and, and Procedures. n.d.). Measurements were carried out in two phases in the frequency range A by use of Bruel&Kajer Type 2236 sound level meter. Measurements were made 1.5 m above ground level with a microphone windscreen, at least 2 m away from buildings to prevent any surface reflection. A-weighted equivalent continuous sound pressure level (LAeq) measurements were made at the roadside measurement point, 3 times on a dry weekday in the winter during the day, evening, and night rush hours. Measured 1-h equivalent continuous sound pressure levels (Leq) reported 78.1 dBA for Lday, 77.1 dBA Levening, and 79.3 dBA for the night. Validation and noise map generations have been made for the daytime period due to low-variance coefficient (1.5%) between the road traffic noise indicators, Lday, Levening, and Lnight.
Then, the daytime measurements have been carried out three times between 07:00-10:00, 12:00-15:00, and 16:00-19:00 at 12 measurement points shown in Fig. 2. The sampling intervals correspond to times of going to work, lunch break, and returning from work, respectively. The measurement time was 15 min so that all measurements could be carried out in peak hour periods within the same day to ensure stability in weather conditions. The results of 15-min measurements and simulation outputs for each point were given in Table 4.
European Commission Working Group Assessment of Exposure to Noise (WG-AEN) specifies the uncertainty in the noise modeling procedure as well as the tolerance for noise mapping outcomes (WG-AEN 2007). The difference between actual measurement results and the calculation outputs reported by the model must not exceed 1 dB at 300 m from the source, 3 dB at 600 m from the source, and 10 dB at 2,000-3,000 m from the source, according to WG-AEN 2006. Table 4 shows that the sound level in 11 of 12 receiver points was higher than the calculated values. The simulation program model input process is the first cause of this discrepancy. The noise modeling technique, according to WG-AEN, causes minor differences between measurement findings and simulation program results. The accuracy level for selected building heights data is 1 dB. In comparison with the smooth asphalt, which was defined in model, as the reference (0 dBA), a correction is needed for ground surface type according to land usage maps. This approach could result in a 1 dBA increase in acoustic uncertainty. As a result, due to the noise modeling procedure, more than 2 dB can be noticed as a reasonable disparity in noise mapping simulation results.
Four of the measurement points, points 1, 2, 3, and 7, do not have another building block between them and the highway, and their distance to the highway is relatively less than the others. It can be seen from Table 4 that the difference between measurement and model results is within the acceptable range at these 4 measurement points where traffic noise is dominant. However, some points are above the tolerable range due to the noise sources that cannot be intervened and affect the background noise such as water elements, children's playground, and construction activities.

Evaluation
In Turkey, The Regulation on Environmental Noise Assessment and Management which was prepared as per the European Directive became effective as of 04.06.2010 upon proclaiming on Official Gazette (European Table 4 Noise measurement results and calculation statistics  Parliament and Council of the European Union 2002) (RENAM 2010). As stated in the regulation, the case study plot is within "Areas with high residential density in zones where commercial buildings and noise-sensitive uses coexist." The noise level should not exceed 63 dBA during the daytime period regarding the regulation. Figure 4 shows the noise levels on the facades in the daytime. Facade noise maps are generated by making calculations for center receiver points on each floor of the building blocks. A total of 580 calculation points have been defined for the facade noise map in the case study site. There are 240 calculation points in the 12 blocks of the first row, where there is no other building block between it and the road. The following inferences have been made by evaluating the calculation outputs of the facade noise map: • 24% of all facades in the settlement; 55% of the facades in the initial row blocks are exposed to noise level above the limit value. • Considering the Southeast facades facing the highway, 39% of all fronts; 88% of the first row facades are exposed to noise levels above the guideline. Since the sleeping units are on the southeastern facade of the blocks, noise control is critical, for the most part of front facades. • When the noise levels are evaluated according to the population distribution in the settlement, 14% of whole residents and 31% of the residents in the first row blocks are exposed to noise level above the limit value.
• The noise level that should not be exceeded according to the regulation is 55 Leq dBA in open areas that are generally used during daylight hours. Accordingly, 44% of the open area was exposed to noise above the threshold value. • When the noise level is evaluated based on floors, 12% of the first floors and 73% of the 5th floors of the building blocks are exposed to the noise level above the limit value.

Simulation of solar physical models and determination of optimum tilt angles
PVsyst was used for the solar part of the study, which includes determining radiation data, energy generation potential, and optimum tilt angles. Calculation of incident radiation in an inclined plane from horizontal light intensity data is defined as the transposition model. PVsyst 6.7.7 offers two transposition models, Hay and Perez (PVsyst Photovoltaic Software n.d.). Perez model was preferred because it gives better results in terms of root mean deviation (RMSD) in calculations, even with synthetic data (Ineichen 2011). This study aims to determine the highest performing noise barrier configuration by investigating the effect of the tilt angles and dimensions of the edges on solar energy and noise control performance. The latitude and the longitude of the case study area are 41°47′ North Latitude and 29°22′ East Longitude, respectively, and its height above mean sea level is 110 m. Since the photovoltaic panels will be integrated into the noise barrier, the area assigns the azimuth of the photovoltaic system. The barrier has been designed between the highway and the settlement, which azimuth is measured 30° counterclockwise from the south. Panels with the same properties were used in all alternatives analyzed in the study. The panel selection allows working with different sizes of the top edge. As a result of market research, a 90 Wp monocrystalline photovoltaic panel was used. The technical details of the panel are given in Table 5.
As seen in Fig. 5, global horizontal irradiation, diffuse horizontal irradiation, and clearness index values, between 56.6-259.9 W/m 2 , 34.4-120.3 W/m m 2 , and 0.34-0.54, respectively. Optimum tilt angles for monthly, seasonal, semi-annual, and annual periods were calculated via the "Optimisation Tool" module of PVsyst 6.7.7 software. Table 6 reveals the determined angles for the area according to certain periods.

Generating PVNB alternatives
In the step of determining the inclination of the double barrier tops, combinations of the annual and other periodic optimum tilt angles are derived. It does not include tilt values below 25° in the calculations as they increase soiling on the PV panels (Lu and Zhao 2018). The annual optimum tilt angle is simulated at the first edge of the barrier top, then at the second edge. Figure 6 shows the flowchart followed during the determination of PVNB alternatives. PVNB alternatives analyzed in the study can be seen in Table 7.

Specifying criteria and criteria weights
The study aims to select the barrier alternative that shows ideal performance in terms of noise control and solar energy. Sixty barrier alternatives with various cross-sections were derived with different diffraction edge widths and periodic optimum inclination angles.
The hybrid system design of solar energy and noise barrier technologies has 2 key criteria. One criterion is the amount of electrical energy generated, which is the performance indicator of the photovoltaic system. The second is the number of receiver points exposed to the noise level above the limit value, which is the noise control performance indicator of the PVNB. As an additional supplementary criterion to the key criteria, the daylight loss caused by the PVNB in the first row buildings is also included in the calculations as environmental impact criteria. Criteria considered in the evaluation are quantitative factors. Criteria weights were determined by the direct rating method.
Alternatives analyzed with TOPSIS (The Technique for Order of Preference by Similarity to Ideal Solution), the multi-criteria decision method widely used in construction projects.
The way followed in weighting the criteria is as below:  Fig. 6 Steps of generating PVNB alternatives Table 7 PVNB alternatives analyzed in the study 1. The main criteria; the number of affected receiver points and electrical energy generation are assumed to have equal weight. 2. Determination of the criteria weight of daylight loss is based on a study in the literature. Besides the physical and biological effects of the noise barriers, Tamura et al. provide a disutility function for an assessment of the preferences of the inhabitants living near the noise barrier and therefore environmentally affected by the project. In the study's scope, the effect of constructing a 3-m-high barrier was evaluated. The three-attribute disutility function created by taking into account the trade-offs between the reduction of environmental effects (noise reduction and reduction in NO 2 ) and the impact of barrier landscape are shown in Fig. 7. Situations represented by different line types in the chart: solid line, there is no countermeasure; the dotted line, decrease of environmental impacts and the landscape obstruction are taken into account; broken line, besides dotted line group, the countermeasures required to mitigate the effect of landscape obstruction (Tamura et al. 1994). In the case study area, the distance of the first row blocks to the highway varies between 20 and 70 m. The disutility function of each building block is determined according to the graph in Fig. 7, and then the calculated mean value was added to column I in Table 8. 3. Relative weights sum rescaled to 1 and calculated values are shown in column III in Table 8.

Selection of optimum PVNB with the MCDA method
Although there are many noise control measures for transport noise, determining the measure is the most critical phase. Because traffic noise is a typical conflict issue between individual mobility needs and a quieter lifestyle (European Union Road Federation, 2004).
There are different decision methods used in determining the noise reduction measure such as cost minimization, cost-effectiveness analysis, cost-benefit analysis, and multicriteria decision analysis.
Multi-criteria decision-making (MCDM) techniques are presented as a powerful tool to help decision-makers select the most sustainable alternative for various construction problems. MCDM methods enable the selection of the optimal one among the alternatives by considering different criteria of a product, problem, or service. They have the advantage of a method that cannot be done in cost-benefit analysis, such as weighting each criterion (Ir Bert Peeters Gijsjan van Blokland 2018). One of the most widely used methods among MCDM methods is the TOPSIS method. It was preferred because of its ease of use, being programmable, and its ability to measure the relative performance of alternatives in a simple mathematical form (Tansel 2012). In the TOPSIS method, the most preferred alternative should be the closest to the positive ideal solution and the furthest from the negative ideal solution (Deng et al. 2000).
As a result of the TOPSIS method, number 20 is determined as the best alternative among the A3 analysis group. The first diffraction edge of the final optimal solution is 3 m wide and has an angle of 58°. The second diffraction edge is 2 m wide and has an angle of 31°. Figure 8 shows the cross-sectional noise map of the ideal PVNB solution and the current situation. Figure 9 shows the outputs of the analyzed alternatives according to the criteria. In the graph, the performance of the A-type barrier is visualized on the left, and the outputs of the B-type barrier are visualized on the right. The horizontal axis represents alternative numbers, and vertical axes represent criteria. "Daylight loss" criterion expresses the radiation loss per unit area on the front facade of the ground Fig. 7 Disutility functions of noise barrier effect for evaluating its environmental impact (Tamura et al. 1994)  floor of the first row blocks, and the low loss is remarkable for the choice of the most appropriate PVNB. "Electricity generation" criterion expresses the amount of annual energy production from the photovoltaic system and high production increases the performance of the barrier. "Decrease in the number of receiving points exposed to the noise level above 63 dBA" criterion refers to the number of receiver points positively affected by the noise barrier and it is aimed to be maximum at making a decision. For example, with the number 10 alternative for all analysis groups (A1, A2, and A3), more receiver points are positively affected by noise reduction, while they show negative performance in terms of energy generation and daylight loss.

Results
When the results of the study are examined, the following results are obtained: • The first alternative from the A3-analysis group shows the best performance in terms of energy generation. The first larger edge of the barrier top has an annual optimum inclination angle (31°) and the other edge has an angle of 27°. In the second alternative of the same analysis group, although both edges have the optimum tilt angle, the energy production potential is lower. This is due to the shading loss caused by the buildings on the opposite side of the road. In urban photovoltaic installations, nearhorizontal panel inclination angles that are lower than the optimum tilt may increase electricity generation as they reduce shading loss. • In the comparison of A-type and B-type alternatives in terms of energy generation criteria, B-type alternatives showed higher performance than A-type alternatives for the same angles. This is related to the fact that a higher position of the edge with the annual optimum inclination angle reduces shading loss in urban applications. • When A-type alternatives are analyzed in terms of energy production, the performance indicator of the A3-analysis group is higher than other analysis groups. Because in the A3-analysis group, the annual optimum tilt angle (31°) is defined to the wider one edge. When B-type alternatives are examined, the same circumstance is observed in the A2-analysis group. • In terms of noise control, the 10th alternative in the A2-analysis group and the 20th alternative in the A3-analysis group show the highest performance. The height of both PVNB types is equal. Planning the wider edge near-vertical angle increased noise control efficiency. • TOPSIS results are shown in Fig. 10. Alternatives with high TOPSIS results are close to the ideal solution. As the planned angle of inclination in PVNB increases, the result values are approaching to the ideal solution. • According to Fig. 10, A2-analysis group A-type barriers and A3-analysis group B-type barriers perform well. These are alternatives where the wide edge is planned with angles close to vertical. • The best alternative determined by TOPSIS also has the best performance in terms of noise control. This shows that noise control is more decisive, although the weights of noise control and solar criteria are equal. This can be explained by the higher variance coefficient of the noise control criterion. Since the change in electricity generation is in a smaller range, the effect on the results is less than the noise control. • Since the edge closer to the road will be more exposed to soiling loss, steeper inclination angle of the panels will positively affect the performance of the system. • The inclination angle of the edges has a negative correlation with solar energy and daylight potential in houses and a positive correlation with noise control. The increasing panel tilt decreases the performance rate of the photovoltaic system, while the energy production is high in panels with the annual optimum tilt angle. A similar situation was observed in the daylight loss on the ground floor of the houses. The high tilt angles reduce the solar radiation on the facade. The PVNB form, which causes minimal daylight loss in buildings, is the 1st-alternative in the A2-analysis group. The lower angle of the longer edge reduced the amount of shading on the facade.
The summarizing correlation matrix showing the relationship between performance indicators and design variables according to noise control and solar performance is shown in Table 9.
− 1 and 1 indicates stronger correlation, red color represents the negative correlation, and blue color represents positive correlation. 0 and closer to 0 cells, indicated in white, means no linear correlation between two variables.
• The correlation is weakest between barrier top inclination and the noise reduction at the 1st-floor level while the correlation between top inclination and 2nd floor is strongest. Although there is a positive correlation between the increase of barrier top tilt and noise reduction at the 3rd, 4 th , and 5th floors, the correlation of tilt is weaker at the 4th and 5th floors. The correlation between d tilt and 2nd floor shows that an increase in the d tilt of the barrier top creates lightly more insertion loss at the 2nd-floor level than in the l tilt. • There is a positive correlation between the barrier top inclination and the noise reduction on the floors, and this correlation is strongest on the 2nd floor and the weakest on the 1st, 4th, and 5th floors, respectively. The correlation between d tilt and 2nd floor shows that an increase in the d tilt of the barrier top creates lightly more insertion loss at the 2nd-floor level than in the l tilt. • A significant relationship between the upper barrier width and the reduction in the noise level on the floors was not observed in this study. • There is a moderate negative correlation between barrier top tilt and the energy generation potential and performance ratio. This negative correlation is greater for l tilt for both performance indicators. For this reason, an edge with a more horizontal angle should be positioned at the upper to increase energy production and performance ratio. • There is a very slight correlation between d, l tilt, and the energy generation potential of PVNB alternatives. This is because all alternatives have the same overall width and nominal power. d width has the positive, and l width has the negative correlation with solar electricity production. Therefore, increasing the width of the d (the edge that is close to the road) increases energy production. Figure 11 shows the relation between the barrier type, analysis group, floor levels, and average noise reduction together. In both barrier types, the noise reduction provided by the barrier alternatives on the floors is maximum on the 2nd floor and minimum on the 5th floor. The results show Table 9 Correlation matrix for design variables and performance indicators GlobInc, incident global irradiation in the collector plane; E_Grid, energy injected into the grid; PR, performance ratio; ShdLoss, irradiance loss due to near shadings; Ill. Loss, illuminance loss on the closest facade due to PVNB; ΔL, noise reduction in dB that there is no significant difference between the noise reduction amounts of the A-type barrier and B-type barrier alternatives. Evaluations show that the noise reduction in the floors is slightly higher when the wider one of the edges has a variable angle. For example, the tilt of l with a width of 3 m in the A2-analysis group ranges from 27° to 58° in the A-type barrier, and the average noise reduction of this type is slightly higher than in the A3-analysis group. This is because a barrier with a higher height performs better in terms of noise control. Among the alternatives that show very close results in the acoustic environment quality of the floors, the option that maximizes solar energy should be considered.

Limitations of the study and future work
It is necessary to mention some constraints and limitations of this study, which should be addressed in future work.
• Although this study assessed technical performance and used comprehensive decision factors to improve the suggested methodology, one of its limitations is that it does not address the economic analysis in evaluating PVNB alternatives. Economic analysis is not covered in this paper. Life cycle costing is not the focus of this study as it takes into consideration overall costs (capital investment, purchase and installation costs, social acceptance costs, operating and maintenance costs, disposal costs) and long-term benefits (electricity sale, CO 2 sale, health cost, increased real estate value, saving cost in building envelope noise control) in a systematic way. It will be the focus of future studies, as partial economic evaluations that do not include overall benefits can lead to misleading results in infrastructure investments. • The 60 PVNB alternatives offered an adequate coverage for the L-shaped road noise barrier with 5-m-wide barrier top over the studied region. A subsequent step would be extending this study to more barrier designs with variable top width. • Environmental impact is defined as daylight loss without considering material properties such as opacity of module and structure, which means that the visible light transmission of material was ignored, which may impact the simulation results. Therefore, the role of opacity in reducing the irradiance loss effect should be considered in future research. • The noise control performance of PVNB alternatives is determined by the amount of receiver points affected by noise reduction for a residential area only. Therefore, insertion loss-oriented evaluation should be considered in future work with different building types and different limit values. • Due to the limitation of site properties, the simulation results do not reflect specific outputs for settlements with blocks higher than 5 floors and with different densities. However, future studies should consider the settlement density and explore the relationship between noise mitigation for buildings higher than 5 floors and PVNB design. • PVNB requires an interdisciplinary approach that provides a combined evaluation of solar energy and noise control technologies. Hybrid technology is analyzed by different software whose input parameters and outputs are independent of each other, since there is not an integrated tool that evaluates all disciplines at the same time.
One of the important limitations of the study is that it does not have a common evaluation scale from ecological, economic, and social perspectives in the analysis of outputs with different units obtained from different software.

Conclusion
The inspiration for the study is the use of noise barriers that protect road adjacent settlements from traffic noise, as a sub-structure for photovoltaic modules. Dysfunctional areas along the roadside can create social and environmental benefits, given the conflict between the need for renewable energy and land scarcity. In this study, it is aimed to make a reasonable selection among PVNB alternatives developed for the reduction of traffic noise and energy generation. In the study, the best

Analysis Group A1
Analysis Group A2 Analysis Group A3 Fig. 11 Average noise reduction on different floor levels according to PVNB design types alternative was selected using the TOPSIS method according to various criteria. Criteria are determined as performance indicators expressing the efficiency of the systems and environmental impact on the settlement. Although different indicators are representing the performance of the photovoltaic system, alternatives were evaluated according to the amount of energy production in the study. Noise barrier effectiveness was assessed according to the number of receiving points exposed to unacceptable noise levels specified in the regulation. Besides, daylight loss due to shadows on the ground floors of the front facades of the first row houses was also considered as a criterion.
Among the steps of the study, the most important two steps are creating PVNB variations with determining optimum tilt angles and specifying the criterion weights.
As a result of the literature review, the double-edge noise barrier, which proved to increase the effectiveness of the sound wall, was studied. In line with the solar radiation data of the region, optimum tilt angles were determined in annual, semi-annual, seasonal, and monthly periods. The height, the total barrier top area, and the annual optimum inclination angle of one edge (31°) remained constant for each barrier, while generating alternatives. Sixty alternatives were derived by changing the positions and widths of the diffraction edges, and also the inclination angle of one edge.
The derived 60 barrier types were tested in three criteria. The direct rating method was used while assigning the weights of the criteria. It is accepted that the criteria determining solar energy and noise control effectiveness have equal weights. In first row buildings, there are trade-offs between maximum noise reduction and loss of landscape view and daylight. An indirect approach has been followed in weighing the environmental impact criterion that takes into account the loss of daylight in buildings. For this, the disutility function developed depending on the distance of the barrier from the buildings was used. The disutility function developed by Tamura et al. is a graphical method that models the mutual concessions of two conflicting factors (Tamura et al. 1994). Criteria weights determined by direct rating and indirect approach were normalized and used in the TOPSIS method.
When the alternative selected as a result of the TOPSIS method is compared with the current situation, it is predicted that the number of receiving points affected by noise will decrease by 44% and annual electricity generation will be 524,804 kWh.
The study provides a useful framework for planning photovoltaic noise barrier installations. The results show that the edges of the roads that adversely affect residential areas are suitable for solar energy gain and photovoltaic noise barriers offer a wide market potential.
In the projection of the expanding transportation network and increasing population density of cities, more studies are needed to take advantage of the production and comfort potential of PVNB technology. Design guidelines that deal with both photovoltaic systems and noise barriers should be prepared. The impact of design variables on cost and barrier effectiveness should also be investigated. The accurate estimation of PVNB's lifetime costs and payback periods will facilitate the future implementation of the system.