DOI: 10.3303/CET25120043 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Paper Received: 16 May 2025; Revised: 27 August 2025; Accepted: 23 October 2025 
Please cite this article as: Lumbera M.P., Encarnacion J.I., 2025, Design Optimization of Small-Scale Horizontal-Axis Wind Turbines Using 
Genetic Algorithm to Minimize CO₂ Emissions, Chemical Engineering Transactions, 120, 253-258  DOI:10.3303/CET25120043 
  

 CHEMICAL ENGINEERING TRANSACTIONS  
 

VOL. 120, 2025 

A publication of 

 

The Italian Association 
of Chemical Engineering 
Online at www.cetjournal.it 

Guest Editors: Bing Shen How, Viknesh Andiappan, Denny K.S. Ng, Hon Loong Lam, Petar S. Varbanov 

Copyright © 2025, AIDIC Servizi S.r.l. 

ISBN 979-12-81206-21-2; ISSN 2283-9216 

Design Optimization of Small-Scale Horizontal-Axis Wind 

Turbines Using Genetic Algorithm to Minimize CO2 Emissions 

 Mikash P. Lumbera*, Job Immanuel B. Encarnacion 

 Department of Mechanical Engineering, University of the Philippines Diliman, Quezon City, Philippines 

 mplumbera1@up.edu.ph 

Small-scale wind turbines are increasingly relevant in the global effort to reduce CO2 emissions while meeting 

growing energy demand, particularly in Philippine off-grid areas with modest energy needs and moderate wind 

speeds. Still, designing efficient turbines for such conditions remains challenging due to trade-offs between 

power output, cost, and emissions. To address these, turbine blade designs must be optimized for better 

performance within environmental and economic constraints. This study introduces a genetic algorithm (GA) to 

optimize horizontal-axis wind turbine blades tailored to Phillippine off-grid areas’ high RPM and low wind speed 

conditions. Improved aerodynamic performance was attained using a Python-based Blade Element Momentum 

(BEM) code, with blade geometry parameterized by a conic equation for chord and twist distribution, controlled 

by an A-coefficient iteratively refined by the GA. Enhanced GA settings—larger parent and population sizes and 

a higher maximum A-coefficient—yielded improved power coefficients (𝐶𝑝) and tip speed ratios (TSR). Although 

initial performance was low under local wind conditions, increasing rotor diameter restored and improved 

efficiency.  The optimized 1-m blade achieved a rated power of 13.91 kW at 4.00 m/s and 496.56 RPM, 

outperforming the base blade designed for 10 m/s. To assess environmental benefits, CO2 emissions were 

estimated by determining blade solidity, scaling for mass, and applying an emissions model. The optimized 

blade exhibited significantly lower CO2 emissions than both the baseline design and a comparable diesel 

generator, underscoring its potential for decentralized renewable energy applications. 

1. Introduction 

The global energy sector is at a critical juncture as the demand for sustainable and low-carbon energy sources 

increases. Wind energy, owing to its abundance and minimal environmental footprint, has emerged as a viable 

alternative to fossil fuels. However, in developing countries like the Philippines, challenges such as high 

electricity costs, frequent power interruptions, and limited access to energy in remote areas. As global warming 

intensifies due to the continued reliance on conventional energy sources, and as energy demand rises alongside 

limited electrification in remote islands, the need for renewable energy solutions grows ever more critical. 

In the Philippines, fossil fuels have continued to dominate the energy mix despite ongoing efforts to harness 

renewable resources over the past years. This highlights a significant opportunity for expanding wind energy 

utilization, particularly through technologies suited for low-wind-speed environments prevalent across the 

archipelago. 

Large-scale horizontal-axis wind turbines (HAWTs), typically with rotor diameters between 80 and 108 m, are 

designed for high-wind-speed regions. This presents a mismatch for the Philippine wind conditions since the 

average wind speeds range from only 4.5 to 7.0 m/s, leading to suboptimal turbine performance and economic 

inefficiency (Silang et al., 2014). Consequently, there is a pressing need to develop small-scale wind turbines 

(SWTs) specifically optimized for low-wind-speed applications. 

Small-scale wind turbines present a practical solution for off-grid electrification in the Philippines, where 

approximately 634 islands remain disconnected from the main grid due to geographic and economic constraints 

(Castro et al., 2021). Unlike large turbines, SWTs benefit from lower production, installation, and maintenance 

costs, making them suitable for rural and isolated communities. 

253



Moreover, economic considerations play a pivotal role in SWT deployment. Downsizing generators and 

simplifying mechanical components can substantially lower manufacturing and operational costs, enhancing the 

feasibility of wind energy systems in resource-constrained settings (Encarnacion et al., 2019). The correlation 

between generator size, cost, and turbine performance further emphasizes the need for design optimization to 

achieve economically viable and environmentally sustainable solutions. 

This study aims to optimize blade design for small-scale HAWTs using a Genetic Algorithm (GA), targeting low-

wind-speed Philippine conditions. Blade performance—heavily influenced by parameters like tip speed ratio 

(TSR), chord length, and twist distribution—critically affects energy capture. Metaheuristic methods like GA have 

grown in prominence in turbine design, due to their capacity to handle complex, multivariable challenges. GA 

mimics evolutionary processes (selection, crossover, mutation) to efficiently probe large design spaces for 

optimal solutions. Recent applications include the work of Muhsen et al. (2020), who optimized a small HAWT 

blade for low-wind-speed conditions using aerodynamic simulations, achieving notable efficiency gains. These 

findings, combined with location-specific wind resource studies such as Gacu et al. (2024), underscore the 

potential for tailored, computationally optimized small-scale wind turbine designs in the Philippines. 

2. Method 

2.1 Validation of the BEM Solver 

Encarnacion et al. (2019) used a Python code to evaluate the performance of the different blade configurations 

using simplified Blade Element Momentum (BEM) Theory. BEM remains an industry-grade tool for preliminary 

wind turbine design. More advanced methods such as Computational Fluid Dynamics (CFD) can capture 

complex flow effects but require substantial computational resources, making them less practical for iterative 

optimization in this work. The same BEM solver was used in this study to calculate different performance 

parameters of a wind turbine. To validate the BEM solver, a case study using the linear base blade of the 

experimental study done by (Abdelsalam et al., 2021) was evaluated using the BEM solver. The generated 

power curve of the BEM solver was compared to the experimental power curve of the base blade. In addition, 

the base blade was evaluated using QBlade, a design software that calculates the performance of s turbine 

design. The chord and twist distributions of the base blade was shown on the figure below.  

2.2 Design Optimization 

Optimizing wind turbine blades was done by altering two parameters only: chord and twist distribution. Both 

distributions were parametrized by the A-value to generate a smooth geometry along the different sections of 

the blade. The blade modification was patterned on the study performed by (Encarnacion et al., 2019) where a 

conic equation, Eq(1), was used to define a distribution.  

𝑟2 + 𝐴𝜆𝑟 + 𝜆2 = 1 (1) 

The distribution defined by the radial coordinate, 𝑟, and the scaling coefficient, 𝜆, were determined by the value 

of A. The graphical pattern of the distribution varies from Elliptical (A<2), Linear (A=2), and Hyperbolic (A>2). 

The implementation of the parametrization reduced the number of variables needed to be optimized. A python 

code was developed to perform the algorithm shown in Figure 1 that optimizes the power coefficient (𝐶𝑝) and 

the corresponding TSR (𝑇𝑆𝑅𝑐𝑜𝑟) through the application of GA on parametrize A-value of both the twist and 

chord distribution.  

 

Figure 1: Optimization Algorithm performed in this Study 

The initial population of 100 individuals was generated by uniformly sampling 10 points for each A-value 

parameter and cross-combining them for both chord and twist distributions. Each individual was evaluated using 

a fitness function based on its aerodynamic performance shown in Eq(2). The Blade Element Momentum (BEM) 

solver calculated the 𝐶𝑝 across a range of tip speed ratios (TSRs), identifying 𝐶𝑝𝑚𝑎𝑥
 and its corresponding TSR. 

These were normalized with respect to the Betz limit and the base blade’s TSR to compute fitness. New set of 

population from parents of the previous generation was updated after every generation where crossover and 

254



mutation are allowed. The GA was coded to end if the set number of generation has been met or the fitness 

function is not changing. 

𝑓𝑖𝑡𝑛𝑒𝑠𝑠 = (
𝐶𝑝𝑚𝑎𝑥

𝐶𝑝𝐵𝑒𝑡𝑧

) × (
𝑇𝑆𝑅𝑐𝑜𝑟

𝑇𝑆𝑅𝐵𝑒𝑡𝑧
) (2) 

2.3 Carbon Footprint Assessment 

To evaluate the environmental performance of the optimized wind turbine blade (WTB), a carbon footprint 

analysis was conducted and compared against both the baseline blade and a diesel generator of equivalent 

rated power. The analysis incorporated real industrial data on commonly used materials for wind turbine blade 

construction, considering two eco-parameters: embodied energy and carbon footprint. These values accounted 

for material production, manufacturing processes, transportation, operational use, and final disposal. The mass 

of the optimized blade was estimated by scaling the results from the study of Algolfilaf et al. (2021), which relates 

blade solidity to mass. The corresponding carbon footprint, in terms of metric tons of CO2 (tCO₂), was then 

estimated using emission factors reported by Morini et al. (2022), based on blade mass. 

For comparison, the baseline blade’s mass was computed using a 3D CAD model in SolidWorks, assuming a 

glass fiber composite material. The same emissions estimation method was applied. Additionally, a 

representative diesel generator with comparable power output was included to benchmark the carbon footprint 

of the optimized WTB against conventional fossil-fueled power generation. 

3. Results 

3.1 BEM Solver Validation 

The BEM solver developed in this study was validated using the linear blade geometry from Abdelsalam et al. 

(2021), which employs the E216 airfoil and a rotor diameter of 0.5 m, optimized for a wind speed of 10 m/s and 

Reynolds number of 100,000. The E216 airfoil was selected for its high lift-to-drag ratio at low Reynolds 

numbers, making it suitable for small-scale applications. 

 

Figure 2: Validation of the BEM solver   

Figure 2 compares the power coefficient curves from experimental data, the custom BEM solver, and QBlade 

with Prandtl’s loss correction. The experimental blade reached a peak. 𝐶𝑝 of 0.425 at TSR = 4.80, the BEM 

solver predicted 0.407 at TSR = 5.75, and QBlade gave 0.385 at the same TSR. The BEM solver’s rightward 

shift in peak performance—linked to its steady-state, inviscid assumptions and lack of turbulence modeling—is 

consistent with Lazzerini et al. (2022). At TSR = 6.25, the BEM solver overestimates power, while QBlade's 

overestimation begins at TSR = 6.62. 

Despite these differences, the BEM and QBlade results show good agreement in curvature from TSR = 5 

onwards. Variations are attributed to the different correction models used: QBlade applies Prandtl’s correction, 

while the custom BEM solver incorporates the iterative solution method for blade element momentum equations 

with guaranteed convergence proposed by Ning (2014), which helps improve accuracy in predicting thrust and 

torque behavior at high angles of attack. 

3.2 Genetic Algorithm’s Performance  

An important observation during the implementation of the genetic algorithm (GA) is the influence of the number 

of parents on the generation at which convergence begins. As shown in Figure 3a, varying the parent count 

significantly affects the convergence behavior of the GA. 

The same parameters validated for the BEM solver were used in the GA, with the exception of a higher Reynolds 

number set to Re = 200,000, to better represent blade sections experiencing flow conditions beyond 100,000. 

255



The number of parents was varied across five cases: 5, 8, 10, 13, and 15. The generations at which convergence 

began were observed at 12, 33, 3, 22, and 6, respectively. An exponential trendline fitted to the data indicates 

that as the number of parents increases, the generation at which the GA converges tends to decrease. 

This behavior is expected due to the increased population size associated with a higher number of parents, 

especially under the complete crossover scheme applied. With more parents contributing genetic material, the 

algorithm explores a broader solution space per generation, thereby accelerating convergence in earlier 

generations. 

  

Figure 3:  Effect of the Number of Parents on Convergence of GA (a), Generated Individuals (b) 

Figure 3b illustrates the distribution of all generated individuals across different parent counts, highlighting the 

first and final (100th) generation, which includes the optimized blade. The results fall within an acceptable design 

region, as no maximum 𝐶𝑝 values occur at extreme TSR values (i.e., TSR = 1–2 and TSR = 7–8), where 

physically unrealistic performance often appears. Additionally, all 𝐶𝑝 values remain below the Betz limit of 0.593, 

further validating the acceptability of the generated solutions. 

A key trend is evident in the population size across different parent counts. As the number of parents increases, 

the number of individuals generated per generation also increases, resulting from the combinatorial crossover 

mechanism used in the algorithm. While this leads to a more extensive exploration of the design space—which 

is beneficial for identifying high-performing blades—it also increases the computational load and run time. 

Notably, with 15 parents, the GA produced a significantly larger population while achieving faster convergence. 

This suggests a trade-off between solution quality and computational cost that must be considered when tuning 

the GA parameters. 

 

Figure 4: Power Curve of Optimized Blades for Different Maximum A-Coefficients (a), Chord (b) and Twist (c) 

Distribution of the Base and Optimized Blade 

The influence of varying the maximum allowable A-coefficient on the aerodynamic performance of the optimized 

wind turbine blade was investigated. The resulting optimized A-coefficients tend to approach the higher values 

of 𝐶𝑝 which attributes to higher fitness of the optimized blades. For a maximum A-coefficient (Amax) of 10, the 

algorithm converged to the chord’s A-coefficient (Ac) value of 8.2. When Amax was increased to 30 and 50, the 

resulting Ac stabilized at 6.7, indicating a threshold beyond which further increases in Amax no longer influence 

the optimized curvature. This saturation suggests that there exists an optimal range of curvature values that 

maximize performance within the given design space. Comparison on the optimized blade’s geometry to the 

base blade is shown in Figure 4b and 4c where both the distributions of the optimized blade are hyperbolic in 

nature. 

The corresponding aerodynamic performance of the optimized blades is illustrated in Figure 4a. Blades 

generated with higher Amax values exhibited superior 𝐶𝑝 curves, especially beyond the TSR of 4.5. Among the 

tested configurations, the blade optimized at Amax = 50 consistently achieved the highest 𝐶𝑝 across a broad 

TSR range. This improvement in performance is particularly beneficial for turbines operating in environments 

a b c 

a b 

256



with variable wind speeds, as it indicates a robust design capable of maintaining high efficiency over a wider 

range of operating conditions. 

Overall, increasing the maximum A-coefficient during optimization enhances both the quality of the geometry 

and the aerodynamic performance, up to a critical point. Beyond this, the gains in fitness and power output begin 

to plateau. The blade optimized at Amax = 50 demonstrated the most favorable characteristics in terms of 

geometry and 𝐶𝑝 distribution, and is thus considered the most effective design among those examined. 

3.3 Carbon footprint assessment 

To analyze the environmental benefits of the optimized wind turbine blade (WTB), a carbon footprint (CFP) 

analysis was conducted and benchmarked against a baseline blade and a diesel generator of comparable power 

output. Using the geometry of the NREL 5 MW reference blade from Algolfilaf et al. (2021), the solidity was 

calculated to be 0.052, and the corresponding total blade mass was estimated at 16,757.1 kg based on its mass 

per unit length distribution. This served as the basis for scaling the mass of the blades generated through the 

genetic algorithm (GA). The scaling relation used was: 

𝑀𝑚𝑜𝑑𝑒𝑙

𝜎𝑚𝑜𝑑𝑒𝑙
=

𝑀𝑔𝑒𝑛

𝜎𝑔𝑒𝑛
 (3) 

Applying this relationship, the estimated mass of the generated blades ranged from 4.01 to 5.08 kg, depending 

on the blade geometry. In comparison, the baseline blade modeled in SolidWorks using glass fiber composite 

materials had a total mass of 5.58 kg. The lower mass of the optimized blade was expected, as the design was 

refined to enhance aerodynamic performance while reducing material use through lower blade solidity. 

The carbon footprint of the blades was then computed using emission factors reported by Morini et al. (2022), 

which correlate blade mass to total CO₂ emissions. The following equation was used: 

𝑀𝑚𝑜𝑑𝑒𝑙

𝐶𝐹𝑃𝑚𝑜𝑑𝑒𝑙
=

𝑀𝑔𝑒𝑛

𝐶𝐹𝑃𝑔𝑒𝑛
 (4) 

Using this approach, the estimated carbon footprint of the GA-generated blades ranged from 0.016 to 0.021 

metric tons of CO₂ (tCO₂). The optimized blade design corresponded to a carbon footprint of 0.017 tCO2. In 

contrast, the baseline SolidWorks model yielded a higher carbon footprint of 0.023 tCO2 due to its greater mass. 

For additional comparison, a 13 kW diesel generator—representing conventional fossil fuel-based power 

generation—from de Melo et al. (2019) was included. Based on published emission factors, the diesel generator 

had an estimated carbon footprint of 0.109 tCO2 for equivalent power output. As shown in Figure 5, the optimized 

blade not only reduced emissions by approximately 22.7 % compared to the baseline blade, but also achieved 

over 84.0 % lower emissions compared to the diesel generator. 

 

Figure 6: Carbon Footprint in tCO2 of the Optimized Blade, Base Blade, and the Diesel Generator  

These findings highlight the significant environmental advantages of the optimized wind turbine blade. With an 

output of 13.91 kW and a minimal carbon footprint, the blade offers a cleaner alternative to traditional fossil fuel 

systems. Furthermore, the reduction in material usage translates to potential cost savings in manufacturing, 

transportation, and end-of-life disposal. Taken together, the optimized blade design demonstrates both 

environmental sustainability and economic viability, reinforcing its applicability in low-emission renewable 

energy systems. 

4. Conclusions 

The application of a Genetic Algorithm (GA) in optimizing horizontal-axis wind turbine blades using an A-

coefficient-based parameterization showed promising results in increasing the blades performance and 

minimizing the blade’s carbon footprint. Increasing the maximum A-coefficient improved 𝐶𝑝 values, particularly 

257



beyond TSR = 4.5, with performance gains plateauing at Amax = 50 — where the optimized blade consistently 

achieved the highest 𝐶𝑝  across a wide TSR range. The convergence of designs to a consistent Ac value 

suggests an optimal curvature zone for efficient energy extraction. The evolution of chord and twist distributions 

further indicated that higher A-coefficients lead to more aerodynamically favorable shapes — characterized by 

reduced chord curvature and enhanced twist near the root. These trends are consistent with physical 

expectations and reinforce the robustness of the GA's search capability. 

From an environmental standpoint, the carbon footprint assessment underscored the optimized blade’s 

sustainability advantage. The computed carbon footprint of the optimized blade (0.017 tCO2) was significantly 

lower than that of the baseline blade (0.023 tCO2) and dramatically less than that of a comparable diesel 

generator (0.109 tCO2). This highlights the viability of GA-optimized wind turbine blades not only in terms of 

aerodynamic efficiency but also in minimizing environmental impact. 

In summary, integrating GA with curvature-controlled design parameters provides an effective framework for 

aerodynamic and environmental optimization. While the present study validates the blade’s aerodynamic 

performance using BEM-based methods, further validation can be conducted using higher-fidelity simulations 

such as (CFD) and experimental testing in a wind tunnel or field conditions. These approaches could serve as 

valuable extensions for future research to confirm and enhance the real-world performance of the optimized 

blade. 

Nomenclature

𝜆 – scaling coefficient 

𝜎𝑔𝑒𝑛  – solidity of blade generated 

𝜎𝑚𝑜𝑑𝑒𝑙 – solidity of model 

𝐶𝐹𝑃𝑔𝑒𝑛 – carbon footprint of blade generated 

𝐶𝐹𝑃𝑚𝑜𝑑𝑒𝑙 – carbon footprint of model 

𝐶𝑝𝐵𝑒𝑡𝑧
 – maximum 𝐶𝑝 of Betz Limit 

𝐶𝑝𝑚𝑎𝑥
 – maximum 𝐶𝑝 of generated blade 

𝑟 – radial coordinate 

𝑀𝑔𝑒𝑛 – mass of generated blade 

𝑀𝑚𝑜𝑑𝑒𝑙 – mass of model 

𝑇𝑆𝑅𝐵𝑒𝑡𝑧 – corresponding TSR of 𝐶𝑝𝐵𝑒𝑡𝑧
 

𝑇𝑆𝑅𝑚𝑎𝑥 – corresponding TSR of 𝐶𝑝𝑚𝑎𝑥
 

Acknowledgments 

The author gratefully acknowledges the University of the Philippines, the Office of the Vice President for 

Academic Affairs (OVPAA), and the Office of the Vice Chancellor for Research and Development (OVCRD) for 

their generous support and grant that enabled the completion of this study. Special thanks are also extended to 

the College of Engineering – High Performance Computing Facility (COE-HPC) for providing the computational 

resources necessary to run the simulations and optimization codes. 

References 

Abdelsalam A.M., El-Askary W.A., Kotb M.A., Sakr I.M., 2021, Experimental study on small scale horizontal axis 

wind turbine of analytically-optimized blade with linearized chord twist angle profile. Energy, 216, 119304. 

Algolfat A., Wang W., Albarbar A., 2022, Dynamic response analysis of a 5 MW NREL wind turbine blade under 

flap-wise and edge-wise vibrations. Journal of Dynamics, Monitoring and Diagnostics, 1(4), 208–222. 

Castro M.T., Ocon J.D., 2021, Can Off-grid Islands Powered by Renewable Energy Microgrids be Operated 

Sustainably without Subsidies? A Techno-economic Case Study in the Philippines. Chemical Engineering 

Transactions, 88, 427-432. 

de Melo F., Silvestre A., Carvalho M., 2019, Carbon footprints associated with electricity generation from 

biomass syngas and diesel. Environmental Engineering and Management Journal, 18(7), 1391–1397. 

Encarnacion J.I., Johnstone C., Ordonez-Sanchez S., 2019, Design of a horizontal axis tidal turbine for less 

energetic current velocity profiles. Journal of Marine Science and Engineering, 7(7), 197. 

Gacu J.G., Garcia J.D., Fetalvero E.G., Catajay-Mani M.P., Monjardin C.E.F., Power C., 2024, A 

Comprehensive Resource Assessment for Wind Power Generation on the Rural Island of Sibuyan, 

Philippines. Energies, 17(9), 2055, en17092055. 

Lazzerini G., Coiro D.P., Troise G., D’Amato G., 2022, A comparison between experiments and numerical 

simulations on a scale model of a horizontal-axis current turbine. Renewable Energy, 190, 919–934. 

Muhsen H., Al-Kouz W., Khan W., 2020, Small Wind Turbine Blade Design and Optimization. Symmetry, 12(1), 

18. 

Morini A.A., Ribeiro M.J., Hotza D., 2021, Carbon footprint and embodied energy of a wind turbine blade—a 

case study. International Journal of Life Cycle Assessment, 26(6), 1177–1187. 

Ning A., 2014, A simple solution method for the blade element momentum equations with guaranteed 

convergence. Wind Energy, 17(9), 1327–1345. 

Silang A., Uy S.N., Dado J.M., Cruz F.A., Narisma G., Libatique N., Tangonan G., 2014, Wind energy projection 

for the Philippines based on climate change modeling. Energy Procedia, 52, 26–37. 

258


	0133.pdf
	Design Optimization of Small-Scale Horizontal-Axis Wind Turbines Using Genetic Algorithm to Minimize CO2 Emissions