Flexible Fuzzy Goal Programming Approach in Optimal Mix of Power Generation for Socio-Economic Sustainability: A Case Study
Abstract
:1. Introduction
1.1. Benefit of Flexible Fuzzy Goal Programming
1.2. Paper Organization
2. Literature Review
3. Methodology
3.1. Multiobjective Optimization Model
3.2. Flexible Fuzzy Goal Programming with Tolerance Functions
3.3. Stepwise Solution Procedure for MOPP
- Step 1:
- Formulate the problem at hand as a multi-objective mathematical model.
- Step 2:
- Obtain the individual optimal solution of the model using any available package, considering one objective at a time.
- Step 3:
- Formulate a pair-wise comparison matrix using the solutions in Step 2, given as:
- Step 4:
- Identify the lowest and highest value of each column in Step 3 obtained from Step 2 above and set them as a lower and upper goal, respectively.
- Step 5:
- Construct the membership functions using the FFGP models in Equation (5).
- Step 6:
- Construct a function that will maximize the overall linear additive model of the auxiliary variables defined from the membership function in Step 5 above.
- Step 7:
- Solve for the function in Step 6 using a suitable optimization package and obtain the goal achievement value.
3.4. Levelized Cost of Energy
3.4.1. Component of levelized Costs
3.4.2. Capital Costs
3.4.3. Operation and Maintenance Cost
3.4.4. Fuel Costs
4. Sustainable Development Goals
4.1. Intended Nationally Determined Contribution (INDC)
- reduce the emissions intensity of its GDP by 33 to 35 percent by 2030 from 2005 levels.
- achieve about 40 percent cumulative electric power installed capacity from non-fossil fuel-based energy resources by 2030, with the help of the transfer of technology and low-cost international finance including from the Green Climate Fund (GCF).
- create an additional carbon sink of to 3 billion tonnes of CO2 equivalent through additional forest and tree cover by 2030.
4.2. Current Scenario of the Indian Power Sector
4.2.1. Installed Capacity of Power Generation
4.2.2. Generation of Electricity
4.2.3. Import and Export of Electricity
4.2.4. Consumption of Electricity
4.2.5. Renewable Energy in India
4.3. State-Wise Scenario of the Power Sector
4.4. Electricity Demand Projections for India
5. Electrical Energy Production Modeling
5.1. Constraints for Electric Energy Production
- Renewable capacity addition targets fixed by Government;
- Must Run Status for Renewable Energy Sources;
- Loss of Load Probability;
- Energy Not Served;
- Provision of Reserve Margin;
- International commitments by the country;
- Emission limits if any; and
- Current Energy mix.
5.1.1. Renewable Capacity Addition Targets Fixed by Government
5.1.2. Must Run Status for Renewable Energy Sources
5.1.3. Loss on Load Probability and Energy Not Served
5.1.4. Provision for Reserve Margin
5.1.5. International Commitments by India
5.1.6. Emission Limits
5.1.7. Current Energy Mix in India
5.2. Mathematical Model Formulation
6. Results Analysis and Discussions
Managerial and Practical Implications
7. Conclusions and Recommendations
- The optimal allocation of installed capacity among different technologies in Table 9 can be adopted to meet the projected demand for electricity;
- For any further increase in demand, the allocation of different electricity plants can be made based on the optimal solution in Table 11;
- Increasing electricity demand should be made by shifting allocations towards renewable energy sources, especially solar, wind and hydro;
- A less-costly decommissioning method can be adopted for plants, especially for the nuclear plant;
- Government should install solar panels on the roof-tops of government offices and encourage individuals to use a solar panel for electricity generation for their personal use;
- Government should reduce the dependency on costly imported coal and explore and use domestic coal reserves for running existing coal plants.
- Shifting from fossil fuels to renewable energy sources may lower carbon dioxide emissions, which furthers climate change mitigation goals (SDG13);
- Ensuring efficient energy access to poorer citizens and deploying large scale renewable sources will positively impact the SDG1 goal of alleviating poverty and all its ramifications;
- Energy efficiency will help achieve the sustainability of cities (SDG11), and with smart cities, road traffic accident risks will reduce drastically, improving peoples’ health (SDG3);
- The provision of access to efficient and affordable energy will create employment opportunities for men and women who will be engaging in hairdressing and digital services (SDG5). It will also improve the quality of education via access to laboratories, internet facilities and modern technologies, helping the necessary flourishing of interdisciplinary research (SDG 10).
Author Contributions
Funding
Acknowledgments
Conflicts of Interest
Abbreviations
COVID-19 | Coronavirus Disease 2019 |
Carbondioxide. | |
SDG | Sustainable Development Goals |
UN | United Nations. |
UNCC | United Nations Conference Centre |
INDC | Intended Nationally Determined Contribution. |
RES | Renewable Energy Sources. |
GOI | Government of India. |
FFGP | Flexible Fuzzy Goal Programming. |
Kwh | Kilo Watt hour |
MWH | Mega Watt Hour |
MU | Million Unit |
GW | Giga Watt |
MW | Mega Watt |
LCOE | levelized Cost of Energy. |
GDP | Gross Domestic Product. |
Kg | Kilo gram |
LOLP | Loss On Load Probability |
ENS | Energy Not Served. |
TERI | The Energy Resources Institute. |
HRES | High Renewables Scenario. |
LRES | Low Renewables Scenario. |
NDRES | Non-Dispatchable Renewable Energy Sources. |
CEA | Central Electricity Authorithy. |
APTEL | Appellate Tribunal for electricity. |
GCF | Green Climate Fund. |
MoSPI | Ministry of Statistics & Programme Implementation. |
MOOP | MultiObjective Optimization Problem. |
DM | Decision Maker(ing) |
MCDM | Multi-Criteria Decision Making |
kth objective function. | |
kth fuzzy aspirational goal. | |
ith resource restriction. | |
coefficient of the jth decision variable in the ith constraint. | |
T | flexible tolerance interval. |
kth goal’s membership value. | |
kth aspirational goal’s lower value. | |
kth aspirational goal’s upper value. | |
an objective function related to optimizing the levelized costs of energy | |
by individual technology. | |
an objective function related to optimizing the levelized costs of | |
energy by all technologies. | |
an objective function related to optimizing the present value of | |
energy by all technologies. | |
Installed Capacity of Hydro. | |
Installed Capacity of Coal. | |
Installed Capacity of gas. | |
Installed Capacity of nuclear. | |
Installed Capacity of solar thermal. | |
Installed Capacity of solar PV. | |
Installed Capacity of Wind. | |
Installed Capacity of Biomass. | |
Installed Capacity of Small Hydro. | |
Initial investment at time t. | |
Operating and maintenance charge at time t. | |
fuel cost at time t. | |
r | discount rate. |
system energy yield at time t. | |
t | time in years. |
References
- Ali, I.; Modibbo, U.M.; Chauhan, J.; Meraj, M. An integrated multi-objective optimization modeling for sustainable development goals of India. Environ. Dev. Sustain. 2021, 2, 3811–3831. [Google Scholar] [CrossRef]
- Karuppiah, K.; Sankaranarayanan, B.; Ali, S.M.; Chowdhury, P.; Paul, S.K. An integrated approach to modeling the barriers in implementing green manufacturing practices in SMEs. J. Cleaner Prod. 2020, 265, 121737. [Google Scholar] [CrossRef]
- D’Adamo, I.; González-Sánchez, R.; Medina-Salgado, M.S.; Settembre-Blundo, D. E-Commerce Calls for Cyber-Security and Sustainability: How European Citizens Look for a Trusted Online Environment. Sustainability 2021, 13, 6752. [Google Scholar] [CrossRef]
- Ahmed, S.; Taqi, H.M.M.; Farabi, Y.I.; Sarker, M.; Ali, S.M.; Sankaranarayanan, B. Evaluation of Flexible Strategies to Manage the COVID-19 Pandemic in the Education Sector. Global J. Flexible Syst. Manag. 2021. [Google Scholar] [CrossRef]
- Giudice, F.; Caferra, R.; Morone, P. COVID-19, the food system and the circular economy: Challenges and opportunities. Sustainability 2020, 12, 7939. [Google Scholar] [CrossRef]
- Ali, I.; Fügenschuh, A.; Gupta, S.; Modibbo, U.M. The LR-Type Fuzzy Multi-Objective Vendor Selection Problem in Supply Chain Management. Mathematics 2020, 8, 1621. [Google Scholar] [CrossRef]
- Mahmud, P.; Paul, S.K.; Azeem, A.; Chowdhury, P. Evaluating Supply Chain Collaboration Barriers in Small-and Medium-Sized Enterprises. Sustainability 2021, 13, 7449. [Google Scholar] [CrossRef]
- The World Bank. 2019. Available online: https://data.worldbank.org/country/india (accessed on 3 September 2019).
- Energy Statistics 2019 (Twenty-Fifth Issue) Central Statistics Office, Ministry of Statistics and Programme Implementation, Government of India New Delhi. Available online: http://mospi.nic.in/sites/default/files/publication-reports/EnergyStatistics2019-finall.pdf (accessed on 3 September 2019).
- Zhang, F. How much do power sector distortions cost to South Asia. In Proceedings of the 2nd IAEE Eurasian Conference, Energy in Eurasia: Economic Perspectives on Challenges, Risks and Opportunities, International Association for Energy Economics, Zagreb, Croatia, 12–14 October 2017. [Google Scholar]
- Devarajan, S.; Nabi, I. Economic growth in South Asia: promising, unequalising, sustainable? Econ. Pol. Wkl. 2006, 41, 3573–3580. [Google Scholar]
- Holmes, F. Unmasking the Asian Giant|Financial Sense. 5 July 2012. Available online: https://www.financialsense.com/contributors/frank-holmes/unmasking-the-asian-giant (accessed on 2 September 2019).
- Andrews, R. Electricity and the Wealth of Nations|Energy Matters. 22 November 2015. Available online: http://euanmearns.com/electricity-and-the-wealth-of-nations/ (accessed on 2 September 2019).
- Burke, P.J.; Stern, D.I.; Bruns, S.B. The impact of electricity on economic development: a macroeconomic perspective. Int. Rev. Environ. Resour. Econ. 2018, 12, 85–127. [Google Scholar] [CrossRef]
- Zhang, T.; Shi, X.; Zhang, D.; Xiao, J. Socio-economic development and electricity access in developing economies: A long-run model averaging approach. Energy Policy 2019, 132, 223–231. [Google Scholar] [CrossRef]
- Central Electricity Authority. Committee on Optimal Energy Mix in Power Generation on Medium and Long Term Basis; Ministry of Power, Government of India: New Delhi, India, 2018.
- Knežević, G.; Topić, D.; Jurić, M.; Nikolovski, S. Joint market bid of a hydroelectric system and wind parks. Comput. Electr. Eng. 2019, 74, 138–148. [Google Scholar] [CrossRef]
- Fekete, K.; Nikolovski, S.; Klaić, Z.; Androjić, A. Optimal re-dispatching of cascaded hydropower plants using quadratic programming and chance-constrained programming. Energies 2019, 12, 1604. [Google Scholar] [CrossRef] [Green Version]
- Kanbur, B.B.; Wu, C.; Fan, S.; Duan, F. System-level experimental investigations of the direct immersion cooling data center units with thermodynamic and thermoeconomic assessments. Energy 2021, 217, 119373. [Google Scholar] [CrossRef]
- Kanbur, B.B.; Kumtepeli, V.; Duan, F. Thermal performance prediction of the battery surface via dynamic mode decomposition. Energy 2020, 201, 117642. [Google Scholar] [CrossRef]
- Kroll, C.; Warchold, A.; Pradhan, P. Sustainable Development Goals (SDGs): Are we successful in turning trade-offs into synergies? Palgrave Commun. 2019, 5, 140. [Google Scholar] [CrossRef] [Green Version]
- Nerini, F.F.; Tomei, J.; To, L.S.; Bisaga, I.; Parikh, P.; Black, M.; Borrion, A.; Spataru, C.; Broto, V.C.; Anandarajah, G.; et al. Mapping synergies and trade-offs between energy and the Sustainable Development Goals. Nat. Energy 2018, 3, 10–15. [Google Scholar] [CrossRef] [Green Version]
- Bisaga, I.; Parikh, P.; Tomei, J.; To, L.S. Mapping synergies and trade-offs between energy and the sustainable development goals: A case study of off-grid solar energy in Rwanda. Energy Policy 2021, 149, 112028. [Google Scholar] [CrossRef]
- Rathi, A.K.A. Promotion of cleaner production for industrial pollution abatement in Gujarat (India). J. Clean. Prod. 2003, 11, 583–590. [Google Scholar] [CrossRef]
- Ghose, M.K. Promoting cleaner production in the Indian small-scale mining industry. J. Clean. Prod. 2003, 11, 167–174. [Google Scholar] [CrossRef]
- Mukherjee, D.P. Barriers towards cleaner production for optimizing energy use and pollution control for foundry sector in Howrah, India. Clean Technol. Environ. Policy 2011, 13, 111–123. [Google Scholar] [CrossRef]
- Pal, P.; Sethi, G.; Nath, A.; Swami, S. Towards cleaner technologies in small and micro enterprises: a process-based case study of foundry industry in India. J. Clean. Prod. 2008, 16, 1264–1274. [Google Scholar] [CrossRef]
- Narayanaswamy, V.; Scott, J.A. Lessons from cleaner production experiences in Indian hosiery clusters. J. Clean. Prod. 2001, 9, 325–340. [Google Scholar] [CrossRef]
- Unnikrishnan, S.; Hegde, D.S. Environmental training and cleaner production in Indian industry—A micro-level study. Resour. Conserv. Recycl. 2007, 50, 427–441. [Google Scholar] [CrossRef]
- Annala, L.; Sarin, A.; Green, J.L. Co-production of frugal innovation: Case of low cost reverse osmosis water filters in India. J. Clean. Prod. 2018, 171, S110–S118. [Google Scholar] [CrossRef]
- Nomani, M.A.; Ali, I.; Fügenschuh, A.; Ahmed, A. A fuzzy goal programming approach to analyse sustainable development goals of India. Appl. Econ. Lett. 2017, 24, 443–447. [Google Scholar] [CrossRef]
- Khatun, F.; Ahamad, M. Foreign direct investment in the energy and power sector in Bangladesh: Implications for economic growth. Renew. Sustain. Energy Rev. 2015, 52, 1369–1377. [Google Scholar] [CrossRef]
- McCollum, D.L.; Echeverri, L.G.; Busch, S.; Pachauri, S.; Parkinson, S.; Rogelj, J.; Riahi, K. Connecting the sustainable development goals by their energy inter-linkages. Environ. Res. Lett. 2018, 13, 033006. [Google Scholar] [CrossRef]
- Hassan, Q.; Garg, N.K. Systems approach for water resources development. Global J. Flexible Syst. Manag. 2007, 8, 29–43. [Google Scholar] [CrossRef]
- Kaur, H.; Singh, S.P.; Glardon, R. An integer linear program for integrated supplier selection: A sustainable flexible framework. Global J. Flex. Syst. Manag. 2016, 17, 113–134. [Google Scholar] [CrossRef]
- Khorasani, S.T. A robust optimization model for supply chain in agile and flexible mode based on variables of uncertainty. Global J. Flex. Syst. Manag. 2018, 19, 239–253. [Google Scholar] [CrossRef]
- Gupta, S.; Fügenschuh, A.; Ali, I. A multi-criteria goal programming model to analyze the sustainable goals of India. Sustainability 2018, 10, 778. [Google Scholar] [CrossRef] [Green Version]
- Modibbo, U.M.; Raghav, Y.S.; Hassan, M.; Mijinyawa, M. A Critical Review on the Applications of Optimization Techniques in the UN Sustainable Development Goals. In Proceedings of the 2021 2nd International Conference on Intelligent Engineering and Management (ICIEM), London, UK, 28–31 April 2021; pp. 572–576. [Google Scholar] [CrossRef]
- Modibbo, U.M.; Ali, I.; Ahmed, A. Multi-objective optimization modeling for analysing sustainable development goals of Nigeria: Agenda 2030. Environ. Dev. Sustain. 2021, 23, 9529–9563. [Google Scholar] [CrossRef]
- Yang, X.; Wang, Y.; Sun, M.; Wang, R.; Zheng, P. Exploring the environmental pressures in urban sectors: An energy-water-carbon nexus perspective. Appl. Energy 2018, 228, 2298–2307. [Google Scholar] [CrossRef]
- Ahmadini, A.A.H.; Modibbo, U.M.; Shaikh, A.A.; Ali, I. Multi-objective optimization modeling of sustainable green supply chain in inventory and production management. Alex. Eng. J. 2021, 60, 5129–5146. [Google Scholar] [CrossRef]
- AlArjani, A.; Modibbo, U.M.; Ali, I.; Sarkar, B. A new framework for the sustainable development goals of Saudi Arabia. J. King Saud Univ.-Sci. 2021, 33, 101477. [Google Scholar] [CrossRef]
- Wang, Y.; Liu, J.; Hansson, L.; Zhang, K.; Wang, R. Implementing stricter environmental regulation to enhance eco-efficiency and sustainability: A case study of Shandong Province’s pulp and paper industry, China. J. Clean. Prod. 2011, 19, 303–310. [Google Scholar] [CrossRef]
- Christainsen, G.B.; Haveman, R.H. The contribution of environmental regulations to the slowdown in productivity growth. J. Environ. Econ. Manag. 1981, 8, 381–390. [Google Scholar] [CrossRef]
- Yabar, H.; Uwasu, M.; Hara, K. Tracking environmental innovations and policy regulations in Japan: case studies on dioxin emissions and electric home appliances recycling. J. Clean. Prod. 2013, 44, 152–158. [Google Scholar] [CrossRef] [Green Version]
- Curtis, E.M.; Lee, J.M. When do environmental regulations backfire? Onsite industrial electricity generation, energy efficiency and policy instruments. J. Environ. Econ. Manag. 2019, 96, 174–194. [Google Scholar] [CrossRef]
- English, J.; Niet, T.; Lyseng, B.; Keller, V.; Palmer-Wilson, K.; Robertson, B.; Rowe, A. Flexibility requirements and electricity system planning: Assessing inter-regional coordination with large penetrations of variable renewable supplies. Renew. Energy 2020, 145, 2770–2782. [Google Scholar] [CrossRef]
- D’Adamo, I.; Gastaldi, M.; Morone, P. The post COVID-19 green recovery in practice: Assessing the profitability of a policy proposal on residential photovoltaic plants. Energy Policy 2020, 147, 111910. [Google Scholar] [CrossRef] [PubMed]
- D’Adamo, I.; Rosa, P. How Do You See Infrastructure? Green Energy to Provide Economic Growth after COVID-19. Sustainability 2020, 12, 4738. [Google Scholar] [CrossRef]
- Hondo, H. Life cycle GHG emission analysis of power generation systems: Japanese case. Energy 2005, 30, 2042–2056. [Google Scholar] [CrossRef]
- Zadeh, L.A. Fuzzy sets. Inf. Control. 1965, 8, 338–353. [Google Scholar] [CrossRef] [Green Version]
- Zimmermann, H.J. Fuzzy programming and linear programming with several objective functions. Fuzzy Sets Syst. 1978, 1, 45–55. [Google Scholar] [CrossRef]
- U.S. Energy Information Administration. Capital Cost Estimates for Utility Scale Electricity Generating Plants. 2016. Available online: https://www.eia.gov/analysis/studies/powerplants/capitalcost/pdf/capcost/assumption.pdf (accessed on 4 September 2019).
- Heptonstall, P. A review of Electricity Unit Cost Estimates. Available online: http://www.ukerc.ac.uk/publications/a-review-of-electricity-unit-cost-estimates.html (accessed on 7 September 2019).
- United Nation Development Programme (n.d) Sustainable Development Goals. Available online: https://www.undp.org/content/undp/en/home/sustainable-development-goals.html (accessed on 7 September 2019).
- Government of India, “Nationally Determined Contirbution,” Working towards Climate Justice. 2016, pp. 1–38. Available online: https://www4.unfccc.int/sites/ndcstaging/PublishedDocuments/India%20First/INDIA%20INDC%20TO%20UNFCCC.pdf (accessed on 3 September 2019).
- Dudley, B. BP Statistical Review of World Energy. 2019. Available online: https://www.bp.com/content/dam/bp/business-sites/en/global/corporate/pdfs/energy-economics/statistical-review/bp-stats-review-2019-full-report.pdf (accessed on 5 September 2019).
- Ministry of Power, Government of India. Guidelines on Cross Border Trade of Electricity-2018. 2018; p. 1. Available online: https://powermin.nic.in/en/content/guidelines-importexport-cross-border-electricity-2018 (accessed on 6 September 2019).
- Tripathi, B. Now, India is the third largest electricity producer ahead of Russia, Japan. Business Standard News. 26 March 2018. Available online: https://www.business-standard.com/article/economy-policy/now-india-is-the-third-largest-electricityproducerahead-of-russia-japan-118032600086-1.html (accessed on 4 September 2019).
- Ali, S. The Future of Indian Electricity Demand: How much, by Whom, and under What Condition? 2018. Available online: https://www.brookings.edu/research/the-future-of-indian-electricity-demand-how-much-by-whom-and-under-what-conditions/ (accessed on 6 September 2019).
- Pathak, S.; Saxena, P.; Ray, A.K.; Großmann, H.; Kleinert, R. Irradiation based clean and energy efficient thermochemical conversion of biowaste into paper. J. Clean. Prod. 2019, 233, 893–902. [Google Scholar] [CrossRef]
- Central Electricity Authority. National Electricity Plan. 2018. Available online: www.cea.nic.in/reports/committee/nep/nep/jan/2018 (accessed on 3 September 2019).
- Central Statistical Organization, MOSPI, Government of India. New Series of National Accounts Statistics. 2010. Available online: :mospi.nic.in/default/files/publication/reports/brochure/2004-05 (accessed on 6 September 2019).
- Bhawan, S.; Puram, R.K. CO2 Baseline Database for the Indian Power Sector; Central Electricity Authority, Ministry of Power, Government on India: New Delhi, India, 2011.
- Srikanth, R. India’s sustainable development goals–Glide path for India’s power sector. Energy Policy 2018, 123, 325–336. [Google Scholar] [CrossRef]
Authors | Optimization Type | Description | Solving Method |
---|---|---|---|
Ali et al. [1] | Multiobjective | modeling India’s SDGs | Classical Goal Programming |
Karuppiah et al. [2] | - | Ranked barriers to implementing greener manufacturing in India | Fuzzy MCDM (DEMATEL, ANP, TOPSIS) |
DAdamo et al. [3]. | - | Ranking e-commerce in European countries amidst pandemic | MCDA and a Likert scale survey |
Ahmed et al. [4] | - | Evaluating strategies for managing COVID-19 in education Sector | Pareto analysis and rough-DEMATEL |
Giudice et al. [5]. | - | Cause-effect analysis of COVID-19 on food security | theme popularity metric. |
Mahmud et al. [7] | - | Evaluating Supply Chain Collaboration Barriers in Small-and Medium-Sized Enterprises. | MCDM (Grey DEMATEL and Fuzzy Best-Worst methods ) |
Zhang et al. [15] | - | Study of the relationship between electricity access and social-economic factors | Bayesian Model Averaging |
Knežević et al. [17] | Biobjective | modeling hydroelectric system and wind parks | Mixed integer linear programming |
Fekete et al. [18] | Biobjective | Addressing congestion problem in the transmission network of hydropower plants | Mixed integer linear programming, Quadratic and Chance-constrained programming |
Mukherjee [26] | - | Evaluation of operational performances of cupola and pollution Control system for optimizing energy use | Descriptive statistics and Factor analysis |
Nomani et al. [31] | Multiobjective | Analysis of the sustainable development goals of India | Fuzzy goal programming |
Khatun and Ahamad [32] | - | Examination of the causal relationship between FDI in the energy and power sector, and economic growth in Bangladesh | Empirical study |
McCollum et al. [33] | - | Study of the interconnectivity of the UN SDGs | Descriptive statistics |
Gupta et al. [37] | Multiobjective | Analysis of India’s economic sectors for sustainable development goals | Fuzzy goal programming |
Modibbo et al. [38] | Multiobjective | modeling and analysis of Nigeria’s SDGs | AHP, Fuzzy goal programming |
AlArjani et al. [42] | Multiobjective | A framework for SDGs in Saudi Arabia | Fuzzy goal programming |
Yang et al. [40] | - | Investigating energy–water–carbon nexus of urban sectors in Shanghai and Beijing. | Environmental input–output model. |
Wang et al. [43] | - | Evaluation of the relationship between environmental regulation and eco-efficiency | De-linking and re-linking tool |
Yabar et al. [45] | - | Study on the impact of environmental policy on technological innovation | Patent data analysis |
Curtis and Lee [46] | - | Study of Onsite industrial electricity generation, energy efficiency and policy instruments | Survey |
English et al. [47] | - | Examining balancing requirements in a decarbonizing electricity system. | Capacity expansion and dispatch model |
D’Adamo et al. [48] | - | An economic assessment of a 3 kW plant in the context of several policy scenarios during a pandemic | Descriptive statistics |
Hondo [50] | - | A life cycle analysis of greenhouse gas emissions from power generation systems | Framework and descriptive statistics |
Present work | Multiobjective | Optimal mix of various technology for electricity generation | Flexible fuzzy goal programming |
Technology | Hydro | Coal | Gas | Nuclear | Solar Thermal | Solar PV | Wind | Biomass | Small Hydro |
---|---|---|---|---|---|---|---|---|---|
Capital Cost (Lakh/MW) | 667.38 | 700.27 | 523 | 1600 | 1200 | 530 | 575 | 605.6 | 846.5 |
Operation & maint. (Lakh/MW) | 27.44 | 20.43 | 28.61 | 32 | 16.8 | 7.42 | 10 | 40 | 29.86 |
Fuel Price (Rs) | 0 | 3 | 5.71 | 4228 | 0 | 0 | 0 | 3.2 | 0 |
Specific Fuel consumption (Kg/Kwh) | 0 | 0.627 | 0.46 | 0.00025 | 0 | 0 | 0 | 1.25 | 0 |
Fuel cost (Lakh/MW) | 0 | 98.86 | 69.02 | 62.03 | 0 | 0 | 0 | 245.3 | 0 |
Capacity factor | 35 | 60 | 30 | 67 | 23 | 19 | 29 | 70 | 45 |
Auxiliary consumption | 1.2 | 5.25 | 5.25 | 7.8 | 1 | 1 | 0.5 | 10 | 1 |
Plant life | 40 | 25 | 25 | 40 | 25 | 25 | 25 | 20 | 35 |
Year | Electrical Energy | Peak Electricity |
---|---|---|
Requirement (MU) | Demand (MW) | |
2021–2022 | 1,566,023 | 225,751 |
2026–2027 | 2,047,434 | 298,774 |
RES | Target RES | RES Installed | Expected RES |
---|---|---|---|
Category | IC as on | Capacity as on | Capacity Addition |
31 March 2022 | 31 March 2017 | from 2017–2022 | |
Solar | 100,000 | 12,289 | 87,711 |
Wind | 60,000 | 32,280 | 27,720 |
Biomass | 10,000 | 8295 | 1705 |
Small Hydro | 5000 | 4380 | 620 |
Total | 175,000 | 57,244 | 117,756 |
Year | Hydro | Nuclear | Gas | Committed RES& Retirement of | |
---|---|---|---|---|---|
(MW) | (MW) | (MW) | Capacity (MW) | Coal (MW) | |
2017–2022 | 6823 | 3300 | 406 | 175,000 | 22,716 |
2022–2027 | 12,000 | 6800 | 0 | 275,000 | 25,572 |
Serial Number | Financial Year | GDP at Constant 2004–2005 |
---|---|---|
Prices (Rupees Crore) | ||
1 | 2004–2005 | 2,971,464 |
2 | 2005–2006 | 3,253,073 |
3 | 2006–2007 | 3,564,364 |
4 | 2007–2008 | 3,896,636 |
5 | 2008–02009 | 4,158,676 |
6 | 2009–2010 | 4,516,071 |
7 | 2010–2011 | 4,918,533 |
8 | 2011–2012 | 5,247,530 |
9 | 2012–2013 | 5,482,111 |
10 | 2013–2014 | 5,741,791 |
11 | 2021–2022 | 9,865,466 |
S/N | Technology | CO2 Emission Factor (tCO2/Mwh) |
---|---|---|
1 | Coal | 0.98 |
2 | Diesel | 0.59 |
3 | Gas | 0.45 |
4 | Lignite | 1.38 |
5 | Hydro | 0.011 |
6 | Nuclear | 0.0242 |
7 | Wind | 0.0295 |
8 | Solar | 0.0534 |
S/N | Technology | Installed Capacity (MW) |
---|---|---|
1 | Coal | 192,163 |
2 | Diesel | 838 |
3 | Gas | 25,329 |
4 | Hydro | 44,478 |
5 | Nuclear | 6780 |
6 | Solar | 12,288 |
7 | Wind | 32,280 |
8 | Bio Power | 8295 |
9 | Small Hydro | 4379 |
S/N | Technology | Capacity | Percentage |
---|---|---|---|
(MW) | Mix (%) | ||
1 | Hydro | 51,301.42 | 11.48 |
2 | Coal | 184,073.40 | 41.17 |
3 | Gas | 26,573.01 | 5.94 |
4 | Nuclear | 10,080 | 2.25 |
5 | Solar Thermal | 35,938.05 | 8.04 |
6 | Solar PV | 64,061.95 | 14.33 |
7 | Wind | 60,000 | 13.42 |
8 | Biomass | 10,000 | 2.24 |
9 | Small Hydro | 5000 | 1.12 |
Total | 447,027.83 | 100 |
S/N | Technology | Generation | Percentage |
---|---|---|---|
(MWH) | Mix (%) | ||
1 | Hydro | 15,550 | 9.73 |
2 | Coal | 917,394 | 57.42 |
3 | Gas | 66,217 | 4.14 |
4 | Nuclear | 54,584 | 3.42 |
5 | Solar Thermal | 71,876 | 4.50 |
6 | Solar PV | 105,631 | 6.61 |
7 | Wind | 151,757 | 9.50 |
8 | Biomass | 55,226 | 3.46 |
9 | Small Hydro | 195,26 | 1.22 |
Total | 1,597,715 | 100 |
Technologies | Tol = 1% | Tol = 2% | Tol = 3% | Tol = 4% | Tol = 5% | Tol = 6% | Tol = 7% | Tol = 8% | Tol = 9% | Tol = 10% |
---|---|---|---|---|---|---|---|---|---|---|
Hydro | 51,686.49 | 51,932.12 | 52,102.44 | 52,227.48 | 52,325.34 | 52,467.24 | 52,651.13 | 52,833.09 | 53,013.37 | 53,192.17 |
Coal | 178,135.4 | 180,485.2 | 183,004.70 | 184,277.40 | 184,138.40 | 183,982.30 | 183,967.60 | 183,953.10 | 183,938.60 | 183,924.20 |
Gas | 26,167.25 | 26,167.01 | 26,167.01 | 26,167.01 | 26,573.01 | 26,167.01 | 26,167.01 | 26,167.01 | 26,167.01 | 26,167.01 |
Nuclear | 10,080 | 10,080 | 10,080 | 10,080 | 10,080 | 10,080 | 10,080 | 10,080 | 10,080 | 10,080 |
Solar Thermal | 80,283.14 | 66,046.33 | 51,809.30 | 40,907.75 | 31,162.40 | 25,697.84 | 24,173.95 | 22,683.91 | 21,223.47 | 19,789.04 |
Solar PV | 20,966.25 | 36,724.26 | 52,629.30 | 65,287.11 | 76,841.70 | 84,029.66 | 87,195.12 | 90,330.46 | 93,439.49 | 96,525.39 |
Wind | 60,450.37 | 59,438.06 | 57,188.34 | 56,389.40 | 56,762.99 | 54,135.70 | 53,103.53 | 52,061.21 | 51,010.02 | 49,951.03 |
Biomass | 8295 | 8295 | 8295 | 8295 | 8295 | 10,000 | 10,000 | 10,000 | 10,000 | 10,000 |
Small Hydro | 5037 | 5061.47 | 5078.07 | 5090.25 | 5099.79 | 5113.62 | 5131.54 | 5149.28 | 5166.85 | 5184.27 |
Tolerance | Electricity Demand | Levelized Cost (lakh/MU) |
---|---|---|
0.01 | 1581683 | 53.02 |
0.02 | 1597343 | 52.46 |
0.03 | 1613004 | 51.95 |
0.04 | 1628664 | 51.52 |
0.05 | 1644324 | 51.16 |
0.06 | 1659984 | 51.17 |
0.07 | 1675645 | 51.12 |
0.08 | 1691305 | 51.07 |
0.09 | 1706965 | 51.03 |
0.10 | 1722625 | 50.98 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Khan, M.F.; Pervez, A.; Modibbo, U.M.; Chauhan, J.; Ali, I. Flexible Fuzzy Goal Programming Approach in Optimal Mix of Power Generation for Socio-Economic Sustainability: A Case Study. Sustainability 2021, 13, 8256. https://doi.org/10.3390/su13158256
Khan MF, Pervez A, Modibbo UM, Chauhan J, Ali I. Flexible Fuzzy Goal Programming Approach in Optimal Mix of Power Generation for Socio-Economic Sustainability: A Case Study. Sustainability. 2021; 13(15):8256. https://doi.org/10.3390/su13158256
Chicago/Turabian StyleKhan, Mohammad Faisal, Asif Pervez, Umar Muhammad Modibbo, Jahangir Chauhan, and Irfan Ali. 2021. "Flexible Fuzzy Goal Programming Approach in Optimal Mix of Power Generation for Socio-Economic Sustainability: A Case Study" Sustainability 13, no. 15: 8256. https://doi.org/10.3390/su13158256
APA StyleKhan, M. F., Pervez, A., Modibbo, U. M., Chauhan, J., & Ali, I. (2021). Flexible Fuzzy Goal Programming Approach in Optimal Mix of Power Generation for Socio-Economic Sustainability: A Case Study. Sustainability, 13(15), 8256. https://doi.org/10.3390/su13158256