0
Research Papers: Energy Systems Analysis

Impact of Air Quality and Site Selection on Gas Turbine Engine PerformanceOPEN ACCESS

[+] Author and Article Information
David W. MacPhee

Department of Mechanical Engineering,
The University of Alabama,
Tuscaloosa, AL 35487
e-mail: dwmacphee@ua.edu

Asfaw Beyene

Department of Mechanical Engineering,
San Diego State University,
San Diego, CA 92182
e-mail: abeyene@mail.sdsu.edu

1Corresponding author.

Contributed by the Advanced Energy Systems Division of ASME for publication in the JOURNAL OF ENERGY RESOURCES TECHNOLOGY. Manuscript received January 30, 2017; final manuscript received September 26, 2017; published online October 17, 2017. Assoc. Editor: George Tsatsaronis.

J. Energy Resour. Technol 140(2), 020903 (Oct 17, 2017) (7 pages) Paper No: JERT-17-1049; doi: 10.1115/1.4038118 History: Received January 30, 2017; Revised September 26, 2017

Abstract

Air pollution can have detrimental effects on gas turbine performance leading to blade fouling, which reduces power output and requires frequent cleanings. This issue is a fairly well-known phenomenon in the power industry. However, site selection for gas turbine installation on the basis of air quality is rarely part of the decision-making process, mainly due to lack of geographical options especially in an urban environment or perhaps due to a simple assumption that air quality at a local micro-level has no impact on the performance of the engine. In this paper, we perform a computational fluid dynamics (CFD) study on an area surrounding a combined heat and power (CHP) facility to assess the impact of local wind distribution on air quality and the performance of a gas turbine engine. Several aerodynamic properties are suggested as possible indicators of air quality and/or high airborne particulate concentration. These indicators are then compared to data collected at various points in and around the site. The results suggest that through post-processing of a simplified CFD simulation analyzing the adjacent terrain, a continuous map of field variables can be obtained and help designers locate future CHP or gas turbine power plants in regions of lower particulate concentrations. This, in turn, would greatly increase efficiency and cost-effectiveness of the proposed power plant.

<>

Introduction

The burning of fossil fuels has increased significantly over the past few decades, and with it, the presence of greenhouse gases in the atmosphere, which have been linked to climate change. In addition, other combustion products, especially when released near population centers, present heath concerns to humans. Due in large part to these health-related interests, there has been a recent surge in research involving pollutant transport modeling, especially for urban areas, to predict air quality and ensure safe air and water are available to the general public.

While such pollutant transport models have garnered a significant increase in research interest lately, many are focused at the micro-scale level and involve, for example, the investigation of pollutant dispersion to ascertain impacts on human health in urban settings [1]. These are often dubbed “street canyon” models, of which there are many examples [26], and have been used effectively for micro-scale dispersion prediction and validation. Although this is certainly an important application of applied computational fluid dynamics (CFD), countless other uses are possible, including, for example, analysis or simulation of pollutant concentration on the performance of gas turbines for power and/or heat generation in urban settings.

Gas turbines present a significant design challenge, especially when one considers the extreme temperatures, pressures, and mechanical loadings experienced in the compression, combustion, and expansion processes. In order to reduce these impacts and increase longevity of gas turbine systems, manufacturers have been continually optimizing system components, for example applying alloy or composite protective coatings [7], which has resulted in a marked increase in turbine longevity over the years. One major problem associated with gas turbine operation—one that cannot be alleviated by simply designing more resistant alloys—is the inevitable decrease in performance over time due to component fouling. Over time, particulate matter, which may result from combustion products or perhaps from pollen, oil, or unburnt fuel [8], can attach to turbine surfaces, which results in steady performance degradation. Due to this inevitable consequence of gas turbine operation, regularly scheduled cleanings must be performed, usually every 2 weeks to 6 months, depending on operating conditions and turbine model specifics. These cleanings range from a simple pressure washing, which can be done while the turbine is actually running, to a complete shutdown and hand-washing of all internal components.

The cleaning process of turbines can be quite costly, especially in cases where more frequent cleanings are required. Incurred costs come either in the form of maintenance costs, which include constant monitoring and on-line cleaning of the turbines where possible, as well as those resulting from lost electricity generation. If power output and fuel usage are monitored, performance degradation on any gas turbine system can be clearly seen. For example, in Fig. 1, actual performance data, in million standard cubic feet per kilo-Watt-hour (kWh), are shown over an entire year for two 5.2 MW gas turbines utilized in a combined heat and power (CHP) plant in San Diego, CA, USA. Data are taken every 15 minutes for the duration of this period, and a “saw-tooth” curve of daily efficiency maxima is quite clearly seen to decrease over time between cleansing events.

The main reason for this loss in performance over time is, as mentioned previously, a result of particulate matter becoming attached to compressor and turbine blades. As a result, pollutant concentration should be a key factor in determining site location for future gas turbines, as areas with higher pollutant concentrations will require more frequent cleanings and result in lower time-averaged system efficiencies, both of which affect operational costs significantly.

While the effects of pollutant concentrations on gas turbine performance have been extensively studied by researchers in the commercial and academic sectors over the years (e.g., Refs. [9] and [10]), surprisingly very little research has been conducted into the role of turbine geographic location, with respect to local ground topology, on the economic viability of a CHP or gas turbine plant.

Due to the nature of fluid flow, air velocity is significantly impacted by local ground curvature, a fact already observed in the wind power industry (e.g., Ref. [11]). Due to the relatively close proximity of gas turbine inlets to the ground, local topology should therefore be taken into account during CHP site selection. Simple hills and valleys adjacent to the proposed CHP site could significantly impact the transport of pollutants into and/or out of the CHP area. The purpose of this study is, then, to investigate the possibility of using a form of a traditional meteorological micro-scale [12] method in predicting the pollutant concentration in and around a gas turbine power plant. More specifically, the aim is to assess the role of traditional fluid dynamics variables (e.g., velocity, pressure, etc.) as opposed to adding extra transport equations (besides the fluid momentum and any other turbulence transport equations), which serves to computationally increase the complexity of simulations.

In order to compare these variables with pollutant concentrations, particulate data are gathered using a P-Trak Particle Counter at strategic points in the area surrounding the aforementioned CHP plant. These concentrations, combined with simulation data, could give considerable insight into possible locations of future gas turbine or CHP plants, increasing power production and economic viability when compared to those sites selected without such measures.

Computational Fluid Dynamics Modeling

To simplify the modeling procedure, consider a rectangular prism, whose six sides correspond to the ground, sky (or, top), sides, inlet and outlet, Fig. 2. For the time being, the exact dimensions of the simulation domain seen in Fig. 2 need not be of concern, as the governing equations and boundary conditions are of sole importance. Also, note that the origin is marked as “O” and the orientation of the Cartesian coordinate system xi is also included for simplicity.

The scale of the problem will be on the micro-scale order, and as a result, the Coriolis force effects have been neglected. Studies of similar scales (e.g., Ref. [13]) have made the same assumption without problem. The equations of motion in the domain will be governed by the steady-state Reynolds-averaged Navier–Stokes turbulence equations. Assuming an incompressible fluid, neglecting gravity and other thermo-physical effects, the mass and momentum conservation equations are written as Display Formula

(1)$∂ui∂xi=0$
Display Formula
(2)$ρuj∂ui∂xj=∂pxi+∂∂xj(2μSji+ρτij)$

where ui is the fluid velocity, Sji is the mean strain rate tensor, and τij is the specific Reynolds stress tensor resulting from applying the foundations of Reynolds averaging. These tensors can be described as follows: Display Formula

(3)$Sji=12(∂uj∂xi+∂ui∂xj)$
Display Formula
(4)$τij=−ui′uj′¯$

Here, the over-bar in Eq. (4) indicates a time average, while the prime symbol indicates a fluctuating quantity. It should be noted that in Eqs. (1)(4), the pressure p and velocity ui are time averaged.

To approximate the specific Reynolds tensor, τij, Eq. (4), turbulence modeling is required. In this paper, the k–ε turbulence model is used, as is similar to other large-scale simulations (e.g., for wind turbine simulations [14,15]). This model makes use of the Boussinesq hypothesis, which relates the Reynolds stress to two other quantities: one, an intuitive variable in turbulence modeling, is the turbulence kinetic energy, k, and the other, is the kinematic eddy viscosity, νTDisplay Formula

(5)$k=ui′ui′¯$
Display Formula
(6)$τij=2νTSij−23kδij$

The standard k–ε turbulence model relates the kinematic eddy viscosity to the turbulence kinetic energy and another variable, the turbulence dissipation, ε. This is termed a two-equation model, since closure can be achieved through the addition of two more transport equations. First, the relationship between the eddy viscosity and the two new turbulence quantities is expressed as follows [16]: Display Formula

(7)$νT=Cμk2ε$

Here, Cμ = 0.09 is a constant. The two transport equations, for k and ε, assuming a steady-state solution can be written as Display Formula

(8)$uj∂k∂xj=τij∂ui∂xj−ε+∂∂xj[(ν+νTσk)∂k∂xj]$
Display Formula
(9)$uj∂ε∂xj=Cε1εkτij∂ui∂xj−Cε2ε2k+∂∂xj[(ν+νTσε)∂ε∂xj]$

Finally, to completely describe the system of equations, the model constants must be defined: Cε1 = 1.44, Cε2 = 1.92, σk = 1.0, and σε = 1.3. These values reflect the typical recommended values for the standard k–ε model.

The solution to this system of equations, given sufficient boundary and initial conditions (to be explained shortly), is accomplished using the SIMPLE (semi-implicit method for pressure linked equations) algorithm, a widely used solver for Navier–Stokes fluids simulations. To completely explain the boundary value problem in question, all that remains is to describe the boundary and initial conditions. First, the velocity boundary is required to mimic a natural average wind speed for the selected site. The boundary in question is oriented to have wind velocity in the x1 direction, Fig. 2. Using a reference wind speed ur in the x1 direction, and the height at which it was measured at, h, a power–law relationship is described, which models the incoming air as a function of the height above ground, or (x3 − e), where e is the elevation in meters, as Display Formula

(10)$u1=ur(x3−eh)a$

Here, a is a constant set to 1/7 or 0.143. This type of velocity profile is often used to calculate power densities of wind, and especially for description of macroscopic wind velocities [17,18].

At all other boundaries, a zero gradient boundary condition is applied normal to the surface Display Formula

(11)$∂ui∂xjnj=0$

where nj is the outward pointing surface normal. Concerning initial conditions, the velocity is set to a value of (ur, 0, 0) m/s everywhere in the domain for the sake of simplicity. The pressure boundaries have similar constraints as the velocity; on the outlet, a pressure of zero is enforced Display Formula

(12)$p=0$

And, on all other boundaries, the zero gradient condition normal to the wall is again invoked Display Formula

(13)$∂p∂xjnj=0$

Due to the nature of the incompressible solver, the units of pressure herein are m2/s2, due to the division by density to facilitate the solution of governing equations. Finally, the pressure is initialized by letting p = 0 everywhere in the domain. Note that since this solver takes into account only relative pressure differences, this zero pressure was chosen arbitrarily.

For the turbulence kinetic energy k, the inlet boundary will have a constant value according to the following: Display Formula

(14)$k=32(urI)2$

where I is the turbulence intensity. The remaining boundaries have the same zero-gradient boundary conditions as the pressure Display Formula

(15)$∂k∂xjnj=0$

Finally, the turbulence dissipation ε at the inlet is assumed constant and defined as follows: Display Formula

(16)$ε=Cμk34l$

where l is the turbulence length scale, taken as the mean side length of volumes used in ground discretization. Other boundaries once again enforce the zero-gradient condition Display Formula

(17)$∂ε∂xjnj=0$
The boundary value problem, now sufficiently described, is solved using openfoam [19], a finite volume toolbox written in C++ and capable of a wide array of multi-physics simulations, including CFD. The domain discretization is completed using blockMesh, a native meshing tool, which creates volumes based on the subdivision of blocks, created from vertices.

In order to subdivide the domain into blocks usable by blockMesh, a program was written that projects terrain data, which contains geocentric latitude ψ, longitude θ, and elevation e from sea level, into data of the form {x1, x2, x3}. If the sample spaces of latitude and longitude are contained in the intervals $[ψmin,ψmax]$ and $[θmin,θmax]$, respectively, then the coordinates of each point in the domain can be constructed according to the arc length experienced at the average latitude and longitude. To adhere to the origin seen in Fig. 2, the mapping of points is described as follows: Display Formula

(18)$x1=R|θ−θmin| cos(ψmax−ψmin2)$
Display Formula
(19)$x2=R|ψ−ψmin|$
Display Formula
(20)$x3=e$

Here, R is taken as the radius of the earth, 6.371 × 106 m. As the domain is assumed rectangular, the maximum height in the x3 direction must also be described, and is taken as a large number, H, which is chosen to be far enough away from the ground so that the upper boundary does not affect the solution Display Formula

(21)$x3,max=H$

As a final note on geometric considerations, it should be pointed out that the previous discussion has assumed that increases in elevation do not affect arc lengths due to their small values when compared to R.

Blocks were then created for each square mapped to the terrain, extending from x3 = e to x3 = H. The subdivision of these volumes is accomplished according to an assumed expansion ratio, Δr, as follows: Display Formula

(22)$Δr=ΔeΔs$

where Δe is the end cell length in the x3 direction and Δs is the beginning cell length. All intermediate cell lengths are interpolated linearly.

Case Study

The area to be considered herein is taken as the region shown in Fig. 3. This domain lies within coordinates ψ ∈ [32.735 deg, 32.81 deg] and θ ∈ [−117.273 deg, −117.05 deg], and represents the area surrounding and upstream of the aforementioned CHP plant at San Diego State University, shown as the red dot in Fig. 3. This area is chosen due to the frequent cleanings required for the gas turbines providing power and waste heat. In fact, the performance data in Fig. 1 are taken directly from measurements from both gas turbines at this particular site.

For this particular case study, the air kinematic viscosity and density are taken as $μ=1.8×10−5kg/(ms)$ and ρ = 1.2 kg/m3, respectively, the turbulence intensity is taken as 1% or I = 0.01, and the reference velocity (see Eq. (10)) is taken as 4.5 m/s, measured at a height of 80 m [21]. The grid discretization results in a projected area of 8.34 km (north-south) by 20.85 km (west-east), and, with 91 × 269 nodes, the average side length gives an estimated turbulence length scale of l = 42.6 m. The ceiling or “top” x3 coordinate in Fig. 2 is chosen to be H = 1.0 × 103 m, and elevation data are extracted from Google Maps [22].

To ensure results are independent of the number of computational volumes, several simulations were conducted with varying mesh sizes while recording the wall shear traction vector, τj, on the ground surface Display Formula

(23)$τj=μni∂uj∂xi$

By calculating the magnitude of global differences in wall traction throughout the domain, the mesh size can be increased gradually until no differences in wall shear maxima and minima, denoted τ* and seen in Eq. (24), are seen. The results of grid independence tests are displayed in Fig. 4

Display Formula

(24)$τ*=(τj,max−τj,min)2$

The final mesh chosen for this study was just over 7.7 × 106 cells, with the first cell center located around 2 cm above ground level. This domain is shown in Fig. 5, along with a visual representation of the ground elevation, associated contours, and block grid discretization. Each block in the domain contains 80 volumes in the x3 direction with a cell expansion ratio of Δr = 500, as well as 2 volumes in each of the x2 and x1 directions.

The simulation was then run using traditional linear second-order, cell-centered finite volume spatial discretization until relative residual errors for the u1, u2, u3, p, ε, and k solvers were less than 10−5. After reaching these tolerances, all flow variables were saved and are analyzed herein. The simulation itself was conducted using a six-core Intel Xeon X5680 processor running at 3.33 GHz with 24GB high-speed RAM, and required just over 3 days to complete in parallel. Parallelization was achieved by decomposing the domain into 3, 2, and 1 sub-domains in the x1, x2 and x3 regions, respectively. Afterward, the domain was reconstructed from sub-domains for post-processing.

In order to compare the CFD results to real data, a series of measurements were taken at specified locations using a P-Trak particle counter, capable of detecting particle sizes from 0.021 μm. At each location, the counter was allowed to measure concentrations for a full day, and the resulting measurements, in particles per cubic centimeter, or pt/cc, are shown in Table 1, along with geographic data for each locale.

Along with the saved variables in this case, i.e., velocity ui, pressure p, turbulence kinetic energy k, and turbulence dissipation ε, the wall shear traction vector magnitude $|τj|$ and pressure gradient magnitude $|∇p|$ were also calculated during post-processing as follows: Display Formula

(25)$|τj|=(τjτj)$
Display Formula
(26)$|∇p|=∂p∂xi∂p∂xi$

These additional variables serve as more factors to compare particle concentrations with the output of a simplified CFD simulation, which will be discussed shortly in Sec. 4. After tolerances were reached and field variables recorded, data for all flow fields were sampled at heights of 0, 2.5, 5, 10, 20, and 40 m above ground level, since CHP gas turbine intakes are generally within this region of elevation above ground. The resulting variables, as compared to measured particulate concentrations, are discussed in Sec. 4.

Results and Discussion

Fluid flow is governed by differences in pressure, so the most intuitive variable to discuss first is the thermodynamic pressure p, which is seen plotted against particle concentration in Fig. 6. Note that the units of p in this case arise from division by the fluid density as is consistent with incompressible flow solvers, but the relative differences remain unaffected. As can be seen in Fig. 6, a general relationship between decreased pressure and decreased concentration is observed. However, this result should be taken lightly, since the pressure gradient is of primary importance in describing fluid flow, not the absolute value. It is more likely then that the pressure gradient magnitude $|∇p|$ may be a better indicator of particulate concentration. The pressure gradient is thus plotted in the same way in Fig. 7. Unfortunately, from assessing Fig. 7, the simulated pressure gradient magnitude is not significantly correlated to the measured particle concentration. There are, however, several other fields readily available from post-processing of results, one of which is the fluid velocity. Much like the pressure gradient, this is a vector quantity but is much easier simplified by assessing its magnitude $|ui|$, defined as follows:

Display Formula

(27)$|ui|=uiui$

The velocity magnitude is directly related to the mass flux, which may be an indicator of the amount of particulate matter transported through a given area. The relationship between velocity magnitude and particle concentration in this simulation is shown in Fig. 8. It is clear that no discernible relationship between velocity magnitude and particulate concentration data of Fig. 8 can be drawn. Although the velocity magnitude should in theory be proportional to the expected particle mass flux (if pollutant generation is constant throughout the domain), this is clearly not the case. This should not be surprising since pollutant generation is not constant in any given domain, and also due to the fact that particulates are often heavier than air, and can settle in regions of low wind velocity (stagnation areas) such as valleys or between mountain ridges. Velocity magnitude, therefore, cannot be relied on as a realistic indicator of particulate concentration.

Due to the nature of the turbulence model used, it is also possible to plot the sampled data of turbulence kinetic energy k and turbulence dissipation ε, which are shown in Figs. 9 and 10.

Obviously, the turbulence kinetic energy k at the ground is assumed to be zero; this is why the data sets corresponding to the ground level are not shown in Fig. 9, in the same way that they were not shown in Fig. 8 due to the no-slip velocity boundary condition. Both the turbulence kinetic energy, Fig. 9, and the turbulence dissipation, Fig. 10, seem to exhibit similar characteristics, which is not at all surprising since turbulence dissipation should in theory be higher in regions of high turbulence kinetic energy.

The wall shear traction vector magnitude from Eq. (25) was also calculated during post-processing and the resulting values are compared to particulate concentration in Fig. 11. Low wall shear could be a general indicator of stagnant pressures areas due to an absence of pressure gradient at that location (and hence, fluid acceleration). As a result, the inverse of wall shear traction vector, $1/|τj|$, is also seen plotted in Fig. 12.

The wall shear traction magnitude inverse seems to comply loosely with the particulate concentration. This is especially apparent when comparing $1/|τj|$ to the concentration at measurement location 5 in Fig. 12; in other words, the simulated continuum fields do predict successfully the region of lowest particulate concentration if inverse shear magnitude at the wall is chosen as the appropriate variable. This suggests that there may be a connection between CFD-simulated wall shear stress and particulate concentration in urban areas as is seen here.

It is clear from Fig. 13 that surface pressure generally decreases along the direction of fluid flow. However, there could be further relationships between pressure gradient, which drives fluid flow and particulate concentration. These regions of low wall shear stress likely represent areas of stagnation, and therefore, more particulate matter is likely to coalesce. For this reason, it is suggested that a low-fidelity CFD simulation, calculating wall shear stress at the ground, could be an excellent predictor of air pollutant concentration, especially those areas which experience low wall shear stress.

Conclusions

Site selection for CHP plants is usually related mainly to factors associated with ease of power and heat distribution in a given area. However, it is well known that the presence of air pollutants can have a detrimental effect on gas turbines, requiring frequent cleanings and associated down-time, which can adversely affect system performance both energetically and monetarily. This study has advocated for a simple CFD approach to better predict areas, which experience low particulate density, better suited to CHP site selection.

The present study has indicated a general approach to modeling a micro-scale simulation using the open-source CFD software, openfoam, and utilizing the standard k–ε turbulence model. Using terrain data and an assumed velocity profile, CFD field variables are compared alongside particulate concentrations, which were collected at strategic locations around a CHP site at San Diego State University, a location where turbine fouling is a major concern and a detriment to turbine performance.

Using the results from CFD simulations, it was found that generally, a high wall shear stress is indicative of low particulate density, and likely a “good” area to host gas turbines or a CHP site. In the case study investigated herein, inverse shear stress vector magnitude was able to determine the best site selection in terms of pollutant concentration. Since pollutant concentration can vary wildly in any urban setting—in this case, some locations measured triple the particle concentration of others—simulations such as this can be a good starting point for CHP site selection as they provide a decent, but more importantly, continuous estimate of air quality at very little cost to the designer. Using data gathered from such simulations, a more efficient site may be selected—one which requires less frequent regular cleansing events. This, in turn, could increase time-averaged turbine output and boost system efficiencies.

Nomenclature

• e =

elevation (m)

• h =

reference height (m)

• H =

domain height (m)

• I =

turbulence intensity (%)

• k =

turbulence kinetic energy (m2/s2)

• l =

turbulence length scale (m)

• ni =

surface normal vector (m)

• p =

thermodynamic pressure [kg/(m s2)]

• R =

• Sij =

mean strain rate tensor (s−1)

• ui =

velocity (m/s)

• ur =

reference wind speed (m/s)

• $ui′$ =

velocity fluctuation (m/s)

• xi =

positional vector (m)

• Δe =

end cell length (m)

• Δr =

cell expansion ratio

• δij =

Kronecker delta

• ε =

turbulence dissipation (m2/s3)

• θ =

longitude (deg)

• μ =

dynamic viscosity [kg/(m s)]

• νT =

turbulence eddy viscosity [kg/(m s)]

• ρ =

density (kg/m3)

• σk, σε, Cμ, Cε1, Cε2 =

turbulence model coefficients

• τij =

Reynolds stress tensor [kg/(m s2)]

• τi =

wall shear traction vector [kg/(m s2)]

• τ* =

difference in maximum and minimum wall shear traction vector [kg/(m s2)]

• ψ =

latitude (deg)

References

Li, X.-X. , Liu, C.-H. , Leung, D. Y. , and Lam, K. , 2006, “ Recent Progress in CFD Modelling of Wind Field and Pollutant Transport in Street Canyons,” Atmos. Environ., 40(29), pp. 5640–5658.
Yamartino, R. J. , and Wiegand, G. , 1986, “ Development and Evaluation of Simple Models for the Flow, Turbulence and Pollutant Concentration Fields Within an Urban Street Canyon,” Atmos. Environ. (1967), 20(11), pp. 2137–2156.
Chan, T. , Dong, G. , Leung, C. , Cheung, C. , and Hung, W. , 2002, “ Validation of a Two-Dimensional Pollutant Dispersion Model in an Isolated Street Canyon,” Atmos. Environ., 36(5), pp. 861–872.
Berkowicz, R. , 2000, “ OSPM—A Parameterised Street Pollution Model,” Urban Air Quality: Measurement, Modelling and Management, Springer, Dordrecht, The Netherlands, pp. 323–331.
Tominaga, Y. , and Stathopoulos, T. , 2011, “ CFD Modeling of Pollution Dispersion in a Street Canyon: Comparison Between LES and RANS,” J. Wind Eng. Ind. Aerodyn., 99(4), pp. 340–348.
Lee, I. , and Park, H. , 1994, “ Parameterization of the Pollutant Transport and Dispersion in Urban Street Canyons,” Atmos. Environ., 28(14), pp. 2343–2349.
DeMasi-Marcin, J. T. , and Gupta, D. K. , 1994, “ Protective Coatings in the Gas Turbine Engine,” Surf. Coat. Technol., 68–69, pp. 1–9.
Meher-Homji, C. B. , Chaker, M. , and Motiwalla, H. , 2001, “ Gas Turbine Performance Deterioration,” 30th Turbomachinery Symposium, Houston, TX, Sept. 17–20, pp. 139–175.
Diakunchak, I. S. , 1992, “ Performance Deterioration in Industrial Gas Turbines,” ASME J. Eng. Gas Turbines Power, 114(2), pp. 161–168.
Kurz, R. , and Brun, K. , 2001, “ Degradation in Gas Turbine Systems,” ASME J. Eng. Gas Turbines Power, 123(1), pp. 70–77.
Hyvärinen, A. , and Segalini, A. , 2017, “ Effects From Complex Terrain on Wind-Turbine Performance,” ASME J. Energy Resour. Technol., 139(5), p. 051205.
Blocken, B. B. , Tominaga, Y. , and Stathopoulos, T. , 2013, “ CFD Simulation of Micro-Scale Pollutant Dispersion in the Built Environment,” Build. Environ., 64, pp. 225–230.
Forthofer, J. M. , 2007, “ Modeling Wind in Complex Terrain for use in Fire Spread Prediction,” Ph.D. thesis, Colorado State University, Fort Collins, CO.
Franco, J. A. , Jauregui, J. C. , and Toledano-Ayala, M. , 2015, “ Optimizing Wind Turbine Efficiency by Deformable Structures in Smart Blades,” ASME J. Energy Resour. Technol., 137(5), p. 051206.
Jackson, R. S. , and Amano, R. , 2017, “ Experimental Study and Simulation of a Small-Scale Horizontal-Axis Wind Turbine,” ASME J. Energy Resour. Technol., 139(5), p. 051207.
Wilcox, D. C. , 2010, Turbulence Modeling for CFD, 3rd ed., DCW Industries, La Cañada, CA. [PubMed] [PubMed]
Blocken, B. , Carmeliet, J. , and Stathopoulos, T. , 2007, “ CFD Evaluation of Wind Speed Conditions in Passages Between Parallel Buildings-Effect of Wall-Function Roughness Modifications for the Atmospheric Boundary Layer Flow,” J. Wind Eng. Ind. Aerodyn., 95(9), pp. 941–962.
Meroney, R. N. , Leitl, B. M. , Rafailidis, S. , and Schatzmann, M. , 1999, “ Wind-Tunnel and Numerical Modeling of Flow and Dispersion About Several Building Shapes,” J. Wind Eng. Ind. Aerodyn., 81(1), pp. 333–345.
Jasak, H. , Jemcov, A. , and Tukovic, Z. , 2007, “ Openfoam: A C++ Library for Complex Physics Simulations,” International Workshop on Coupled Methods in Numerical Dynamics, Dubrovnik, Croatia, Sept. 19–21, pp. 1–20.
NREL, 2017, “ National Renewable Energy Laboratory: California—Average Annual Wind Speed at 80 m,” National Renewable Energy Laboratory, Golden, CO, accessed Oct. 11, 2017,
View article in PDF format.

References

Li, X.-X. , Liu, C.-H. , Leung, D. Y. , and Lam, K. , 2006, “ Recent Progress in CFD Modelling of Wind Field and Pollutant Transport in Street Canyons,” Atmos. Environ., 40(29), pp. 5640–5658.
Yamartino, R. J. , and Wiegand, G. , 1986, “ Development and Evaluation of Simple Models for the Flow, Turbulence and Pollutant Concentration Fields Within an Urban Street Canyon,” Atmos. Environ. (1967), 20(11), pp. 2137–2156.
Chan, T. , Dong, G. , Leung, C. , Cheung, C. , and Hung, W. , 2002, “ Validation of a Two-Dimensional Pollutant Dispersion Model in an Isolated Street Canyon,” Atmos. Environ., 36(5), pp. 861–872.
Berkowicz, R. , 2000, “ OSPM—A Parameterised Street Pollution Model,” Urban Air Quality: Measurement, Modelling and Management, Springer, Dordrecht, The Netherlands, pp. 323–331.
Tominaga, Y. , and Stathopoulos, T. , 2011, “ CFD Modeling of Pollution Dispersion in a Street Canyon: Comparison Between LES and RANS,” J. Wind Eng. Ind. Aerodyn., 99(4), pp. 340–348.
Lee, I. , and Park, H. , 1994, “ Parameterization of the Pollutant Transport and Dispersion in Urban Street Canyons,” Atmos. Environ., 28(14), pp. 2343–2349.
DeMasi-Marcin, J. T. , and Gupta, D. K. , 1994, “ Protective Coatings in the Gas Turbine Engine,” Surf. Coat. Technol., 68–69, pp. 1–9.
Meher-Homji, C. B. , Chaker, M. , and Motiwalla, H. , 2001, “ Gas Turbine Performance Deterioration,” 30th Turbomachinery Symposium, Houston, TX, Sept. 17–20, pp. 139–175.
Diakunchak, I. S. , 1992, “ Performance Deterioration in Industrial Gas Turbines,” ASME J. Eng. Gas Turbines Power, 114(2), pp. 161–168.
Kurz, R. , and Brun, K. , 2001, “ Degradation in Gas Turbine Systems,” ASME J. Eng. Gas Turbines Power, 123(1), pp. 70–77.
Hyvärinen, A. , and Segalini, A. , 2017, “ Effects From Complex Terrain on Wind-Turbine Performance,” ASME J. Energy Resour. Technol., 139(5), p. 051205.
Blocken, B. B. , Tominaga, Y. , and Stathopoulos, T. , 2013, “ CFD Simulation of Micro-Scale Pollutant Dispersion in the Built Environment,” Build. Environ., 64, pp. 225–230.
Forthofer, J. M. , 2007, “ Modeling Wind in Complex Terrain for use in Fire Spread Prediction,” Ph.D. thesis, Colorado State University, Fort Collins, CO.
Franco, J. A. , Jauregui, J. C. , and Toledano-Ayala, M. , 2015, “ Optimizing Wind Turbine Efficiency by Deformable Structures in Smart Blades,” ASME J. Energy Resour. Technol., 137(5), p. 051206.
Jackson, R. S. , and Amano, R. , 2017, “ Experimental Study and Simulation of a Small-Scale Horizontal-Axis Wind Turbine,” ASME J. Energy Resour. Technol., 139(5), p. 051207.
Wilcox, D. C. , 2010, Turbulence Modeling for CFD, 3rd ed., DCW Industries, La Cañada, CA. [PubMed] [PubMed]
Blocken, B. , Carmeliet, J. , and Stathopoulos, T. , 2007, “ CFD Evaluation of Wind Speed Conditions in Passages Between Parallel Buildings-Effect of Wall-Function Roughness Modifications for the Atmospheric Boundary Layer Flow,” J. Wind Eng. Ind. Aerodyn., 95(9), pp. 941–962.
Meroney, R. N. , Leitl, B. M. , Rafailidis, S. , and Schatzmann, M. , 1999, “ Wind-Tunnel and Numerical Modeling of Flow and Dispersion About Several Building Shapes,” J. Wind Eng. Ind. Aerodyn., 81(1), pp. 333–345.
Jasak, H. , Jemcov, A. , and Tukovic, Z. , 2007, “ Openfoam: A C++ Library for Complex Physics Simulations,” International Workshop on Coupled Methods in Numerical Dynamics, Dubrovnik, Croatia, Sept. 19–21, pp. 1–20.
NREL, 2017, “ National Renewable Energy Laboratory: California—Average Annual Wind Speed at 80 m,” National Renewable Energy Laboratory, Golden, CO, accessed Oct. 11, 2017,

Figures

Fig. 1

Performance data for the two 5.2 MW turbines over a calendar year. Large jumps in performance indicate turbine cleansing events.

Fig. 2

Simplified computational domain

Fig. 3

Terrain under investigation, with simulation area shown in box. Circle indicates position of gas turbines. Adapted from Google Earth [20].

Fig. 4

Results of mesh size independence tests. Note: vertical axis shown in log scale.

Fig. 5

Discretized domain, showing (a) entire domain, (b) elevation, (c) elevation contours, and (d) ground grid spacing

Fig. 6

Pressure at various specified heights above ground as compared to particle concentration

Fig. 7

Pressure gradient magnitude at various specified heights above ground as compared to particle concentration

Fig. 8

Velocity magnitude at various specified heights above ground as compared to particle concentration

Fig. 9

Turbulence kinetic energy at various specified heights above ground as compared to particle concentration

Fig. 10

Turbulence dissipation at various specified heights above ground as compared to particle concentration

Fig. 11

Wall shear stress magnitude at various specified heights above ground as compared to particle concentration

Fig. 12

Inverse of wall shear stress magnitude at various specified heights above ground as compared to particle concentration

Fig. 13

Wall shear traction magnitude (top) and pressure (bottom) at the ground

Tables

Table 1 Measurement locations and measured particle concentrations

Discussions

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related Proceedings Articles
Related eBook Content
Topic Collections