Introduction

In street canyons, the dispersion of vehicle emissions is an important topic in urban air quality. Many studies have been conducted to analyse the influence of various parameters on air flow and pollutants dilution inside street canyons. Atmospheric conditions and urban layouts are macroscopic parameters to influence the urban ventilation and pollutant dispersion capacity1,2,3,4,5. As for street canyon in urban city, so much parameters have been investigated to study their effect on the air flow and pollutant dilution such as street aspect ratios (building height/street width)6, ambient wind speed and directions7, buoyancy force induced by wall heating or solar shading8,9,10, canyon configuration11, different roof combinations12, building height variations13, lift-up building design14, tree planting15,16, viaduct settings17, advertisement boards18 etc.

Solar PV panels are usually fixed on building roof in many Chinese cities for power generation purpose and the effect of PV panels on wind pattern and pollutant distribution have been study in the past years. There are studies indicate that PV panels can reduce urban temperatures and mitigate the heat island effect19,20. Huang et al.21 shown that the effect of PV installation on ventilation varies by position. Tian et al.22 verified that PV roof and PV facade with ventilated air gap significantly change the building surface temperature and sensible heat flux density. Wang et al.23 examined the impact of PV panels on the internal temperature, lighting conditions and building energy consumption. Similar study was conducted by Zhong et al.24 shown that semi-open street roofs have an overall negative impact on pollutant dispersion.

In the research methods of flow field and air pollutant diffusion, CFD simulation is a useful methodology with high temporal and spatial resolution. Reynolds-Averaged Navier-Stokes (RANS) approaches are most commonly adopted to predict the flow and pollutant dispersion in urban models, although large eddy simulations (LES) have been confirmed to have higher accuracy while they need much more computing resources. Among the RANS models, the RNG k–ε model prediction accuracy is better in simulating weak-wind regions of urban districts such as the sheltered region behind the buildings. The RNG k–ε model is one of the widely-used RANS models, which has been validated to provide reliable simulations for the urban flow25,26,27. Many studies indicated the superior performance of the RNG k–ε model in building simulations28,29, among which there are many studies also focused on the street canyon. Chan et al. found the RNG k–ε model outperformed standard and realizable k–ε model in the prediction of the concentration field in a 2D street canyon model30. By comparing the simulation results of the street canyon model with a series of aspect ratios to the experiment data, Koutsourakis et al. found that the RNG k–ε model had better performance for predicting the cavity flow among popular RANS models31. Therefore, the widely-used commercial CFD software Ansys Fluent with RNG k–ε model is employed for our study.

As mentioned above, although there is already a large body of researches on the impact of solar panels on urban microclimate, building energy consumption, pollutant dispersion and wind velocity. However, in the existing literatures, PV panels are either placed horizontally on rooftops or installed vertically on building facades. There has been little research on the impact of solar panels tilted at specific angles and mounted on rooftops.Therefore, the main purpose of this paper is to investigate the impacts of solar PV panels installed at a certain angle on the roof on the flow field and pollutant distribution in street canyon.

Methodology

Numerical model

The governing equations for the flow and turbulent quantities of RNG \(\:\text{k}-{\upepsilon\:}\) model are as below:

Continuity equation:

$$\:\begin{array}{c}\frac{\partial\:{u}_{j}}{\partial\:{x}_{j}}=0\end{array}$$
(1)

Momentum equation:

$$\:\begin{array}{c}\frac{\partial\:{u}_{i}}{\partial\:t}+{u}_{j}\frac{\partial\:{u}_{i}}{\partial\:{x}_{j}}=-\frac{1}{\rho\:}\frac{\partial\:p}{\partial\:{x}_{i}}+\frac{\mu\:}{\rho\:}\frac{{\partial\:}^{2}{u}_{i}}{\partial\:{x}_{j}\partial\:{x}_{j}}-\frac{\partial\:}{\partial\:{x}_{j}}\left(\overline{{{u}_{i}^{{\prime\:}}u}_{j}^{{\prime\:}}}\right)+{g}_{i}\end{array}$$
(2)

k and \(\:{\upepsilon\:}\) transport equations in the standard \(\:\text{k}-{\upepsilon\:}\) turbulence model:

$$\:\begin{array}{c}\frac{\partial\:k}{\partial\:t}+{u}_{j}\frac{\partial\:k}{\partial\:{x}_{j}}=\frac{1}{\rho\:}\frac{\partial\:}{{x}_{j}}\left[\right(\mu\:+\frac{{\mu\:}_{t}}{{\sigma\:}_{k}}\left)\frac{\partial\:k}{\partial\:{x}_{j}}\right]+\frac{{G}_{k}}{\rho\:}-\epsilon\end{array}$$
(3)
$$\:\begin{array}{c}\frac{\partial\:\epsilon\:}{\partial\:t}+{\mu\:}_{j}\frac{\partial\:\epsilon\:}{\partial\:{x}_{j}}=\frac{1}{\rho\:}\frac{\partial\:}{{x}_{j}}\left[\right(\mu\:+\frac{{\mu\:}_{t}}{{\sigma\:}_{\epsilon\:}}\left)\frac{\partial\:\epsilon\:}{\partial\:{x}_{j}}\right]+\frac{1}{\rho\:}{C}_{\epsilon\:1}{G}_{k}\frac{\epsilon\:}{k}-{C}_{\epsilon\:2}\frac{{\epsilon\:}^{2}}{k}\end{array}$$
(4)

The species transport equation:

$$\:\begin{array}{c}\frac{\partial\:{C}_{i}}{\partial\:t}+{\mu\:}_{j}\frac{\partial\:{C}_{i}}{\partial\:{x}_{j}}=\frac{1}{\rho\:}\frac{\partial\:}{{x}_{j}}\left[\right({D}_{i}+\frac{{\mu\:}_{t}}{S{c}_{t}}\left)\frac{\partial\:{C}_{i}}{\partial\:{x}_{j}}\right]\end{array}$$
(5)

Where

$$\:\begin{array}{c}-\:\overline{{{u}_{i}^{{\prime\:}}u}_{j}^{{\prime\:}}}=\frac{1}{\rho\:}{\mu\:}_{t}(\frac{\partial\:{u}_{i}}{\partial\:{x}_{j}}+\frac{\partial\:{u}_{j}}{\partial\:{x}_{i}})-\frac{2}{3}k{\delta\:}_{ij}\end{array}$$
(6)

\(\:-\:\overline{{{u}_{i}^{{\prime\:}}u}_{j}^{{\prime\:}}}\) is the Renolds stress, \(\:{\mu\:}_{t}=\frac{\rho\:{C}_{\mu\:}{k}^{2}}{\epsilon\:}\) is the turbulent viscosity. \(\:{\upmu\:}\) is the molecular viscosity; \(\:{g}_{i}\) is the gravitational body force; \(\:{G}_{k}\) is the turbulent kinetic energy production; \(\:{\sigma\:}_{k}\) and \(\:{\sigma\:}_{\epsilon\:}\) are the turbulent Prandtl numbers for \(\:\text{k}\) and \(\:{\upepsilon\:}\); \(\:{C}_{i}\) is the pollutants concentration; \(\:{D}_{i}\) is the diffusivity; \(\:S{c}_{t}\) is the turbulent Schmidt number. \(\:{\delta\:}_{ij}\) is the Kronecker delta whose value is 1 when \(\:\text{i}=\text{j}\) and 0 otherwise.

Model descriptions and boundary conditions

This study considers idealized 2D street canyons with a simplified urban geometry when the length of the street canyon is significantly greater than its width and approaching wind pendicular to the street axis32,33,34. Figure 1(a) and (b) show the schematic of base 2D street canyon model and the grid layout of three representative street canyon. The scale ratio of the physical model to the full-scale model is 1:10. The building height (H) of the CFD models is 3 m. The width (W) of the target street canyon and building width(B) are also set as 3 m. A line source that 0.1 m wide continuously emitted CO at a constant flow rate was located at the center of the canyon floor.

The local schematic of four target street canyon models are shown in Fig. 1(c). In this study, three representative installation angles(30°, 45°, and 60°) were selected to investigate the influence of the solar PV panels on the velocity and pollutant distribution within the street canyon. We define case [30], case [45] and case [60] to represent the model with installation angle of the solar PV panels 30°, 45°, and 60° respectively. The length of solar PV panels is set to 0.15 m in model corresponding to a full-scale length of 1.5 m.

The grid size of first layer was 0.002 m and the grid expansion ratio is 1.05 from wall surfaces toward the upper zone. The near-wall area was resolved by the enhanced wall functions directly on the condition that the y + of the first near-wall mesh was less than 5 and the total number of grid cells is 851,992.

The symmetry and outflow boundary conditions were applied at the domain top and outlet, respectively.

At the domain inlet, a power-law velocity profile, turbulent kinetic energy and dissipation rate were used as followed:

$$\:\begin{array}{c}{U}_{in}={{U}_{ref}\left(\frac{y-{y}_{ref}}{{y}_{ref}}\right)}^{\alpha\:}\end{array}$$
(7)
$$\:\begin{array}{c}{{k}_{in}\left({U}_{in}\left(y\right)\ast\:{I}_{in}\right)}^{2}\end{array}$$
(8)
$$\:\begin{array}{c}{\epsilon\:}_{in}\left(y\right)=\frac{{C}_{\mu\:}^{\raisebox{1ex}{$3$}\!\left/\:\!\raisebox{-1ex}{$4$}\right.}{k}_{i}^{\raisebox{1ex}{$3$}\!\left/\:\!\raisebox{-1ex}{$2$}\right.}}{\kappa\:y}\end{array}$$
(9)

Where, \(\:{\upalpha\:}\)=0.22, C = 0.09,\(\:{I}_{in}=0.1\),\(\:{\upkappa\:}=0.41\), \(\:{y}_{ref}=H=3m\)

The Reynolds numbers (\(\:\frac{{{\uprho\:}\text{u}}_{\text{r}\text{e}\text{f}}\text{H}}{{\upmu\:}}\)) are 64,285 which is more than 11,000 to ensure Re number independence35.

The governing equations are discretized using the finite volume method and pressure velocity couplings as achieved via the SIMPLE algorithm. Convective terms in governing equations are discretized using a second-order upwind scheme and diffusive terms employ a central difference scheme. The residuals reached the following minimum values or less than \(\:{10}^{-5}\)for all variables which including continuity equation, velocity components \(\:k\), \(\:\epsilon\:\:\)and pollutant concentration.

Fig. 1
figure 1

a Simulated street canyon model. b Grid arrangements of local region. c Schematic of four target canyon models.

Full size image

Grid independence

A grid independence study was tested on the base model: coarse grid (the minimum grid size was 0.005 m), medium grid (the minimum grid size was 0.002 m), and fine grid (the minimum grid size was 0.001 m). All grid generation had the same grid expansion ratio of 1.05. Figure 2 compares the horizontal velocity profiles along the vertical centerline of the street canyon. The velocity profiles demonstrate good agreement between the medium and the fine grid results and the coarse grid is significantly different compared to medium and the fine grid. The medium grid was choosed in the study.

Fig. 2
figure 2

Grids independence analysis.

Full size image

Evaluation of CFD model

CFD validation of flow modelling

We adopted the wind-tunnel experiment model using by Allegrini to validate the simulation accuracy36. Figure 3 show wind tunnel model and grid layout of street canyon. Since the wind direction is perpendicular to street canyon and the length is much larger than the width of the street canyon, a 2D computational domain is build.

The near-wall grid was refined to 0.001 m and grid expansion ratio was set to 1.05.The total number of grid cells is 47,569. At the outflow boundary, a zero-gradient condition was prescribed. At the top of the domain, a symmetry condition was applied. As for ground and building surfaces, non-slip were imposed for the airflow and pollutant dispersion respectively. The measured velocity and TKE at the vertical centerline position of the street canyon in reference were used as inlet boundary condition in the simulation.

We selected the RNG \(\:\text{k}-{\upepsilon\:}\) model with enhanced wall functions in this study. The residuals reached the following minimum values or less than \(\:{10}^{-5}\) for all variables. The results were compared with the wind tunnel tests under isothermal condition of \(\:{u}_{ref}\)=1.45 m/s and \(\:{R}_{e}\) =19,200. The results shown in Fig. 4 indicated that the model have a good performance and agree with the measured data for wind speed, while the turbulent kinetic energy was slightly under-predicted near the windward wall side for horizontal centerline. The root mean square error(REMS) of \(\:{u}_{x}\),\(\:{u}_{y}\),turbulent kinetic energy(\(\:k\)) on horizontal center line and turbulent kinetic energy(\(\:k\)) on vertical center line are 0.19,0.23,0.47 and 0.39 respectively.

Fig. 3
figure 3

a Wind tunnel model. b Grid arrangement.

Full size image
Fig. 4
figure 4

Validation profiles obtained from CFD simulations and wind tunnel data. a The comparison of normalized horizontal velocity and TKE between numerical simulations and experimental data on horizontal center line. b The comparison of normalized vertical velocity and TKE between numerical simulations and experimental data on vertical center line.

Full size image

CFD validation of dispersion modelling

The CFD results are compared with the wind tunnel data by Meroney37. Figure 5 shows the schematic diagram and arrangements. The size of the street canyon are the same as in wind tunnel experiments. There are 20 uniform street canyons upstream from the target street canyon and 8 street canyons downstream. The building height H and street wide W were set as 0.06 m and the total height of the computational model is 1 m. A line source continuously released a mixture of ethane and air was placed along the center of the street floor to mimic vehicular emissions (\(\:{\text{Q}}_{\text{e}}\)=4 L/h, \(\:{\text{Q}}_{\text{air}}\text{=100l/h})\). Measurements are taken of the vertical profiles of tracer gas concentration along the windward and leeward wall surfaces. Since the wind direction is perpendicular to an isolated street canyon and the length is much larger than the width of the street canyon, a 2D computational domain is considered. The minimum gird size is 0.02 mm at the wall surfaces and grid expansion ratio was set to 1.05. The total number cells is 772,579. The residuals reached the following minimum values or less than \(\:{10}^{-5}\) for all variables.

In the numerical simulation of this validation case, the inflow profiles of the horizontal wind velocity was specified to match the corresponding wind tunnel experimental data, which were given as follows:

$$\:\begin{array}{c}u\left(z\right)={\text{u}}_{\text{r}\text{e}\text{f}}ln{\left(\frac{\text{z}-{\text{d}}_{0}}{\text{h}-{\text{d}}_{0}}\right)}^{{\upalpha\:}}\end{array}$$
(10)

Where \(\:{\text{d}}_{0}\) is the displacement height which is about 2 mm; h is the boundary layer height; \(\:\text{u}\left(\text{z}\right)\) the velocity at elevation z, and \(\:{\text{u}}_{\text{r}\text{e}\text{f}}\) is the velocity at the boundary layer height. The vertical wind profile exponent \(\:{\upalpha\:}\) was estimated to be 0.28.

The concentration is presented in dimensionless form as:

$$\:\begin{array}{c}K=\frac{\text{C}{\text{U}}_{\text{r}\text{e}\text{f}}\text{H}\text{L}}{{\text{Q}}_{\text{e}}}\end{array}$$
(11)

Where C is the volume fraction of ethane, \(\:{\text{U}}_{\text{r}\text{e}\text{f}}\) is wind velocity measured in the free stream at 0.60 m above the tunnel floor which was set to 3 m/s in the study. H is the height of building(0.06 m), L is line source length(0.9 m) and \(\:{\text{Q}}_{\text{e}}\) is the source emission rate(4 L/h).

Figure 6 presents a comparison between the simulated and measured dimensionless pollutant concentration K, distributions on the windward and leeward walls of the canyon. It can be observed clearly that the CFD results agree quite well with the wind tunnel observations for all measurement positions on both the windward and leeward walls of the canyon. Therefore, the current CFD model is feasible for simulating turbulent wind flow and pollutant dispersion inside isolated street canyons.

Fig. 5
figure 5

a Model configurations of wind tunnel experiment. b Mesh schematic of local region.

Full size image
Fig. 6
figure 6

Comparison of K between wind tunnel experiment and CFD simulation.

Full size image

Results and discussions

Impacts of solar panel on velocity in canyon

Figure 7 depicts the wind velocity magnitude distribution in street canyon for all study cases. It can been found that leeward-side wind speed is much smaller than windward-side in all cases due to outside airflow enters the canyon from the windward side and then gradually decelerates. Solar PV panel make an impact on velocity distribution in street canyon compared with the base model.

As shown in Fig. 7, we can found that case [30] has the greatest impact on wind field in all three models with solar PV panel. For case [30], low-speed area in the central of street canyon become smaller and higher wind speed zone increase around the vortex especially in the lower part of the street canyon compared with the base case. The effect of solar PV panel on wind velocity gradually weakens with the installation angle increases. As depict in Fig. 7(c) and (d), the the low-speed zone increases and the high-speed zone decreases compared with the case [30]. No matter what, ventilation condition is still better in case [45] and [60] than the base case.

This phenomenon can be attributed to two reasons. Firstly, The wind speed increases as air flows over the solar panel due to the Bernoulli principle. The smaller the installation angle of the solar panel, the more pronounced the Bernoulli effect becomes, resulting in higher wind speeds at the top of the PV. And so, higher-speed air enters the street canyon from the windward side leading to higher wind speed at the leeward side and larger higher air velocity area in the canyon. Secondly, low-pressure zone forms behind the PV panel due to the combined effect of air viscosity, inertia and Bernoulli principle at the top of leeward wall side. The low-pressure zone draw more air upward and out of the street further increased the wind speed in the canyon. But the vertical height of the solar panel decreases as the installation angle increases that weakening the Bernoulli effect and further decreasing wind speed. Besides this, low-pressure area shrinks as the installation angle increases that makes it more difficult for air to flow out of the canyon.

Fig. 7
figure 7

Velocity magnitude contour in canyon street a base case, b case [30], c case [45], d case [60].

Full size image

Impacts of solar panel on flow matter in canyon

Figure 8 shows the streamline distribution in street canyon for all study cases. Overall, air flows downwards along the windward side and upwards along the leeward side and a apparent clockwise vortices formed in canyon street in all study cases. Two corner vortices are formed in the lower part of the street canyon. Compared to the base case, the notable difference is the the main vortex streamline extends out of the street canyon to a certain extent in the cases with solar PV panels, particularly in case [30].

The reasons of this phenomenon is similar to the explanation in the section “Impacts of solar panel on velocity in canyon”. The combined action of air viscosity and inertia and Bernoulli effect results in the formation of a low-pressure zone behind the PV when air flows around it. As we all know that the larger the frontal area of the obstacle, the larger the area of the low-pressure zone behind it under the same wind speed condition. The low-pressure zone on the top of leeward wall make it easier for upward air to flow out of the canyon that lead to streamlines extend out of the canyon. As shown in Fig. 8, case [30] has the maximum PV vertical height meaning the biggest frontal area that creates the largest area of low pressure followed by case [45] and case [60].

Fig. 8
figure 8

Streamlines in canyon street a base case, b case [30], c case[45], d case[60].

Full size image

Impacts of solar panel on pollutant distribution in canyon

Figure 9 displays the pollutant distribution in canyon street for various study cases. Pollutant concentration is much higher on the leeward side than the windward side in all study cases. The main reason is pollutants move towards the leeward-side wall due to the clockwise vortex and accumulate at the lower-left corner region within the street canyons. Furthermore, small corner vortices near the leeward side further enhance pollutant accumulation. It can also be found that the solar PV pane exert remarkable influence on the pollutant distribution in street canyon. Compared with the base case, the low-concentration region in case [30] expands and shifts to the left, while the area of high-concentration regions is reduced. As the installation angle increases to 45°, the low-concentration area decreases and the high-concentration area increases compare to case [30]. In case [60], the low-concentration region further diminishes. Overall, the base case exhibits the largest area of high pollutant concentration, followed by case [45], case [60] then case [30].

The primary reasons for this phenomenon are wind speed and airflow patterns are two key factors affecting pollutant dilution in street canyons.

High wind speeds indicate good ventilation capacity which is conducive to the dilution of pollutants. As illustrated in the Fig. 7, the largest area of high wind speed occurs in case [30] compared to the other three scenarios which make pollutants more easy to diffusion. The air velocity gradually decreases with the installation angle increasing which weaken the ventilation capacity in the canyon. Base case experiences the lowest wind speed leading to the highest pollutant concentration. As for airflow patterns in Fig. 8, the area covered by streamlines extending beyond the canyon is the largest in case [30]. We can infer that the exchange of pollutants with outside of street canyon is the most vigorous in this scenario which lead to more pollutants are transported out of the canyon. It can be found from Fig. 8 that the extent of this streamline extension decreases with the installation angle increases and results in high pollutant concentration, especially for the base case.

Fig. 9
figure 9

Dimensionless pollutant concentration in canyon street a base case, b case [30], c case [45], d case [60].

Full size image

Mean dimensionless pollutant concentration in street canyons

To quantitatively compare the pollutant concentrations in the street canyon, we calculated the average dimensionless pollutants concentration of the canyon in each case. Figure 10 shows the average dimensionless pollutant concentration (K) in street canyons for the different cases. The mean K value was 81.9, 64.1, 65.6 and 70.7 for base case, case [30], case [45] and case [60] respectively. The mean K value is lowest in case [30] which was 64.1 and decreased by 21.7% compared to the base case. The mean K value for case [45] and case [60] decreased by 19.9% and 13.6% respectively compared to the base case. The reason can be attributed to that have been discussed in Sect. 4.3. The results mean that solar PV panels can improve ventilation condition in the street canyon and the actual effectiveness depend on the installation angle of solar PV panel.

Fig. 10
figure 10

Mean dimensionless pollutant concentration in a street canyon.

Full size image

Impacts of solar panel on pollutant concentration along the building surface

The pollutant concentration near building surfaces can be used to as an indicator of indoor air quality. To further quantify the influence of solar PV panels on pollutant distribution along building walls, we compared the concentrations near the leeward and windward walls under various study cases, as shown in Fig. 11. It can be found that the concentration on the leeward wall is significantly higher than that on the windward wall, with both decreasing toward upper-level regions of the walls.

For the windward wall, the highest pollutant concentration occurs in the base case. The other three study cases show minimal differences in concentration, which can be attributed to two following reasons. First, the area near the windward wall belongs to a high wind speed zone, which promotes pollutant diffusion. Second, pollutants emitted from the road surface are transported toward the leeward side due to the clockwise vortex.

For leeward wall, the lowest concentration on is observed in case [30], followed by case [45] and then case [60]. The highest concentration occurs in the base case. As discussed in the section “Impacts of solar panel on velocity in canyon” and “Impacts of solar panel on flow matter in canyon”, case [30] has the best ventilation capacity compared to other cases. One important factor is the upward wind speed near the leeward wall is greatest in case [30]. Another key factor is that the biggest area of streamline extend out street canyon occurred in case [30], especially in the vicinity of the leeward side. Both of the two factors facilitates the removal of pollutants from the canyon leading to the lowest pollutant concentration. The effect of both factors gradually diminish with the installation angle of PV increasing.

Another noteworthy phenomenon is the rapid decrease in air pollutant concentration at the roof level of the street canyon in the base case, whereas relatively high concentrations persist near the rooftop region in the other three cases. This difference can be attributed to the strong shear forces generated at the roof level in the absence of solar PV panels, which enhance pollutant dispersion and reduce concentration. In contrast, the presence of panels significantly weakens the shear force in the upper right corner of the canyon, resulting in poorer pollutant dispersion and higher concentration levels.

Fig. 11
figure 11

Dimensionless pollutant concentration near windward and leeward wall.

Full size image

Impacts of solar panel on pollutant concentration at pedestrian level

Figure 12 shows the distribution of pollutant concentration at pedestrian level (z = 0.15 m in models) for all study cases.

The highest pollutant concentration occurs at about x/W = 0.3, which is about 0.5 m downwind of the emission source. This phenomenon can be attributed to the transport of pollutants by wind from the windward to the leeward wall at pedestrian level. But, the air can not carry pollutant more far due to the small wind speed at this height. We can also found that solar PV panel with installation angle 30° has the greatest impact on pollutant concentration. The reason is that case [30] has greatest wind speed at this height. The concentration on the leeward side is significantly higher than that on the windward side. This is primarily due to the main clockwise vortex transporting pollutants toward the leeward side, while small corner vortices at the Leeward side within the street canyon further promote the accumulation of air pollutants. It can be observed that the base case experience highest pollutant concentration across the entire pedestrian height, followed by case [60]. A slight difference is noticeable between case [30] and case [45] near the leeward wall, with a moderately higher concentration in case [45]. There is almost no difference on the windward side. The possible reason is that almost all pollutant were carried to the leeward wall side due to the clockwise vortex. Pollutants at lower right corner are primarily caused by diffusion.

Fig. 12
figure 12

Dimensionless pollutant concentration at pedestrian level height.

Full size image

Limitations

It is worth mentioning that, urban models are fairly simplified and 2D models were investigated. The results will become more complicated if we take into account the actual street model and unsteady condition. Further more, more installation scenarios of PV panels should be taken into consideration.

Conclusions

Three installation angles(30°, 45°, and 60°) of PV panels on the top of building were investigated the influence of the PV panels on the velocity and pollutant distribution within the street canyon. Solar photovoltaic panels present on the top of building modify the wind and flow field inside the canyon and generally reduces pollutant concentration in the canyon. PV panels can increase wind speed in street canyon and the effect diminishes as the installation angle increases of PV panels. The flow field in the street canyon changes in the cases with solar panels, especially near the top of the canyon. The streamline extend out from the street canyon due to the present of solar panel that facilitates the exchange of pollutants between the canyon interior and outside. The smaller the installation angle, the more pronounced the effect. Solar panels installed at an angle on rooftop can reduce the pollutant concentration in the canyon and the effectiveness decreases as the angle increases. Base on the study, to maximize the improvement in pollutant dispersion within the street canyon, the primary recommendation is a PV panel installation angle of 30°.

Although further investigations are still required to provide practical guidelines, this paper is one of the attempts to quantify how PV panels setting influence flow and pollutant exposure in 2D urban models, which can provide effective methodologies and meaningful references to urban planning.