Forecasting coal power plant retirement ages and lock-in with random forest regression

•Historical evidence could help identify factors influencing the retirement of CFPPs•We identified most important ages•We forecast some countries to face considerable hurdles retiring plants Coal power are single largest contributor global heating. With new constructions slowing globally, there is an urgent need accelerate existing plants. numerous barriers hampering this, speeds differ considerably across countries. It thus unclear at what age will retire and where hotspots early or late retirements may be located. Our machine-learning model identifies plant-level CO2 emissions penetration renewables in electricity mix as ages. Daring a look into future brings good news: likely earlier than historical average 40 years, generating large reduction benefits. However, we difficulties due high lock-in, influenced by plant ages dependence on coal. These merit targeted support international community. Averting dangerous climate change requires expediting coal-fired (CFPPs). Given multiple here world’s CFPPs. We use supervised machine learning first learn from past, determining that retirements. then apply our dataset 6,541 operating under-construction units 66 Based results, also associated carbon degree which locked coal power. Contrasting with roughly years over 2010–2021, forecasts for 63% current CFPP units. This results 38% less if assuming trends. lock-in index CFPPs power, capacity number units, young continues exacerbate air pollution.1IEAElectricity Market Report.2022Google Scholar Bucking efforts phase-out expedite (CFPPs), have recently built plants.2Jewell J. Vinichenko V. Nacke L. Cherp A. Prospects powering past coal.Nat. Clim. Chang. 2019; 9: 592-597https://doi.org/10.1038/s41558-019-0509-6Crossref Scopus (101) Google Scholar,3Casey J.A. Su J.G. Henneman L.R.F. Zigler C. Neophytou A.M. Catalano R. Gondalia Chen Y.T. Kaye Moyer S.S. et al.Coal-fired closures retrofits reduce asthma morbidity local population.Nat. Energy. 2020; 5: 398-408https://doi.org/10.1038/s41560-020-0600-2Crossref PubMed (18) In addition 6,500 totalling 2,000 GW exist another 450 construction risks expand this fleet.4Global Energy MonitorGlobal Plant Tracker.2022Google Once built, typically operate between 50 years.5Farfan Breyer Structural changes generation towards sustainability risk stranded investments supported indicator.J. Clean. Prod. 2017; 141: 370-384https://doi.org/10.1016/j.jclepro.2016.09.068Crossref (138) Scholar,6Shearer Fofrich Davis S.J. Future proposed India.Earth’s Futur. 408-416https://doi.org/10.1002/2017EF000542Crossref (77) Scholar,7Erickson P. Kartha S. Lazarus M. Tempest K. Assessing lock-in.Environ. Res. Lett. 2015; 10: 084023https://doi.org/10.1088/1748-9326/10/8/084023Crossref (105) If all achieved similar lifetime, 278 Gt (Gt-CO2) would result (calculated authors based Global Monitor (2022)). consume more half remaining budget (500 Gt-CO2) permitted scenarios limit post-industrial warming 1.5°C.8IPCCMitigation Climate Change Summary Policymakers (SPM). Cambridge Univ. Press, 2022: 1-30Google parallel, pollution damage human health prolonged decades.9Koplitz S.N. Jacob D.J. Sulprizio M.P. Myllyvirta Reid Burden disease rising southeast Asia.Environ. Sci. Technol. 51: 1467-1476https://doi.org/10.1021/acs.est.6b03731Crossref (95) New developments slowed.10Littlecott Roberts Şenlen Ö. Burton Joshi Shearer Ewen No 2021: The Collapse Pipeline.2021Google 2021, newly installed declined 2015 levels, when biggest installation occurred (see Figure S1; Table S1). Yet remains unsolved challenge how tackle fleet government interventions wealthy developing economies8IPCCMitigation Scholar,11Edenhofer O. Steckel J.C. Jakob Bertram Reports coal’s terminal decline exaggerated.Environ. 2018; 13: 024019https://doi.org/10.1088/1748-9326/aaa3a2Crossref (82) knowledge conditions affect Many obstacles expected impede CFPPs.12Steckel political economy coal: lessons learnt 15 country case studies.World Dev. Perspect. 2021; 24100368https://doi.org/10.1016/j.wdp.2021.100368Crossref (9) Scholar,13Jakob Political Economy Coal. Routledge, 2022https://doi.org/10.1016/0301-4215(86)90099-6Crossref Scholar,14Diluiso F. Walk Manych N. Cerutti Chipiga Workman Ayas Cui R.Y. D. Song al.Coal transitions - Part 1: systematic map review study learnings regional, national, experiences.Environ. 16: 113003https://doi.org/10.1088/1748-9326/ac1b58Crossref (21) can arise resistance industry, whom trigger financial losses,15UNFCCCJust Transition Workforce Creation Decent Work Quality Jobs.2016Google Scholar,16Manych Why coal? – sector Philippines.Energy Sustain. 62: 113-125https://doi.org/10.1016/j.esd.2021.03.012Crossref (10) but difficulty provoking economic societal disturbances power-generation coal-extraction sector.17Caldecott B. Clark Koskelo Mulholland E. Hickey Stranded assets: environmental drivers, challenges, supervisory responses.Annu. Rev. Environ. Resour. 46: 417-447https://doi.org/10.1146/annurev-environ-012220-101430Crossref (6) addition, many lack technological capacity18Sen Ganguly Opportunities, issues renewable energy development discussion.Renew. 69: 1170-1181https://doi.org/10.1016/j.rser.2016.09.137Crossref (261) Scholar,19Shukla A.K. Sudhakar Baredar Renewable resources South Asian countries: policy recommendations.Resour. 3: 342-346https://doi.org/10.1016/j.reffit.2016.12.003Crossref immediately shift alternative technologies. Strategies needed confront these financial, political, social “lock in” thereby prolonging lifetime emissions.7Erickson Scholar,20Trencher G. Downie Hasegawa Asuka Divestment trends Japan’s businesses.Renew. 124109779https://doi.org/10.1016/j.rser.2020.109779Crossref (20) Several studies attempted fleet. But research limited top-down approaches, volumes schedules prescribed accord mitigation scenarios.21Kefford B.M. Ballinger Schmeda-Lopez D.R. Greig Smart fossil fuel deep decarbonisation scenarios.Energy Pol. 119: 294-306https://doi.org/10.1016/j.enpol.2018.04.018Crossref (40) Scholar,22Oberschelp Pfister Raptis C.E. Hellweg emission generation.Nat. 2: 113-121https://doi.org/10.1038/s41893-019-0221-6Crossref (115) Scholar,23Maamoun Kennedy Jin X. Urpelainen Identifying retirement.Renew. 126109833https://doi.org/10.1016/j.rser.2020.109833Crossref (25) Scholar,24Cui Hultman Edwards M.R. He Sen Surana McJeon H. Iyer Patel Yu al.Quantifying operational lifetimes under Paris goals.Nat. Commun. 10 (4759–9)https://doi.org/10.1038/s41467-019-12618-3Crossref Moreover, usefulness estimates reduced tendency governments market actors ignore budgets deciding schedules. Meanwhile, employing bottom-up approach those modeling plant-by-plant conditions.25Cui Bowman Clarke al.A strategy high-ambition phaseout China.Nat. 12: 1468https://doi.org/10.1038/s41467-021-21786-0Crossref Conversely, although recent scholarship adds understanding lifetimes,26Brutschin Schenuit Van Ruijven Riahi Exploring enablers ambitious phaseout.PaG. 2022; 200-212Crossref it still future. Consequently, anticipate currently CFPPs, result, experience therefore aim determine results. estimate forecasted quantify To achieve extracted literature descriptions influence collected county-specific data publicly available sources (World Bank [2018],27World BankCarbon Price. World Bank, 2018: 50-51https://doi.org/10.4324/9781315270326-31Crossref BP [2021],28BPBP Statistical Review 2021.2021Google EMBER [2021],29EMBERGlobal Electricity etc.). Next, used measure ages, examining 1,697 retired 2010 2020 34 subsequently resulting emissions. Finally, develop share mix, total capacity. makes three contributions. First, country-specific conditions, propose ensuing Second, identifying retirements, generate hints Third, novel meriting attention policymakers investors because their challenges phasing out section presents forecasting, estimated emissions, describes sources, sample constructions, calculation procedures. following discussion, final details experimental academic led identification 1 13 country-level availability inclusion analysis drivers retirement. Country-level describe socio-economic, environmental, governance specific year, whereas reflect characteristics unit examined along sources).Table 1Country-level retirementFactorsData availabilityIndicator unitSourceTime-series availabilityLiteratureAssumed influenceNotesIncluded analysisF1: emissionsyesmillion tons CO2/yearGlobal (2022)4Global ScholaryesTrencher al.,30Trencher Healy Discursive electricity: narratives regime.Energy 132: 782-796https://doi.org/10.1016/j.enpol.2019.06.020Crossref (52) Hughes al.31Hughes J.P. Regulatory treatment uneconomic plants.Electr. 2016; 29: 28-32https://doi.org/10.1016/j.tej.2016.07.008Crossref (1) ScholarPlants higher older less-efficient Due requirements vulnerability regulations, prioritize polluting plants.Annual indicate absolute volumes, unit’s capacity, heat rate, type combusted. rate calculated combustion technology (i.e., subcritical, supercritical, ultra-supercritical), well capture storage (CCS).F2: Carbon priceyesyes/noWorld (2018)27World ScholarnoTrencher Mo al.,32Mo Duan Quantifying implied newly-built become asset pricing.Energy Econ. 99105286https://doi.org/10.1016/j.eneco.2021.105286Crossref (23) Pahle al.,33Pahle Germany’s dash exploring factors.Energy 2010; 38: 3431-3442https://doi.org/10.1016/j.enpol.2010.02.017Crossref (55) Ross,34Ross M.T. industry: implications taxes.Energy 73: 393-409https://doi.org/10.1016/j.eneco.2018.03.022Crossref (8) al.35Mo Zhang W. Tu Q. Yuan Fan Y. Pan Meng Z. role national pricing China’s power.iScience. 24102655https://doi.org/10.1016/j.isci.2021.102655Abstract Full Text PDF (36) ScholarCarbon price policies increase cost generation, reducing competitiveness against other sources.Countries indicated possessing during 2010–2020 receive score “1” “0” not.F3: priceyesUS$/tonBP (2021)28BPBP al.33Pahle ScholarA increases costs, competitiveness. low prices competitiveness.Average Northwest Europe, US Central Appalachian, Japan steam CIF.F4: access rateyes% population electricityWorld (2021)36World BankEnvironmental, Social Governance (ESG).2021Google ScholaryesManych Jakob16Manych ScholarLow electrification rates create supply. choose they established reliable technology, supply stable baseload power.Training set: Since unavailable only one 2019 has not reached 100% (India); assume reach India already 98% 2019.Model application available, extrapolate access.F5: Reliability supplyyes1–8World (2021)37World BankPrice Electricity. Nimble fins, 2021Google reliability drive effort provide electricity. such attractive power.Since insignificant 2018 2019.F6: GDP per capitayesUS$/capitaWorld ScholaryesScholvin,38Scholvin Africa’s policy: constrained nature path dependency.J. Afr. Stud. 2014; 40: 185-202https://doi.org/10.1080/03057070.2014.889361Crossref (12) Hao Tan al.39Tan Thurbon Sung-Young Kim J.A.M. Overcoming incumbent clean shift, How act agents station closure China.pdf.Energy Policy. Crossref (15) ScholarPoor cheap sources. costs encourage delay retirements.F7: policyyes1–100World (2021)40World BankRegulatory Indicators Sustainable Energy.2021https://rise.esmap.org/Google ScholaryesGallagher al.41Gallagher K.S. Bhandary Narassimhan Nguyen Q.T. Banking Drivers demand Chinese overseas Bangladesh, India, Indonesia Vietnam.Energy 71101827https://doi.org/10.1016/j.erss.2020.101827Crossref (33) ScholarPolicies deployment, decreasing time, pushing mixes costs.Based pillar “renewable energy” (RISE).Since data.F8: effectivenessyes1–100Wendling al., (2020)42Wendling Z.A. Emerson J.W. de Sherbinin Esty D.C. al.Environmental Performance Index. Yale Center Environmental Law & epi.yale.edu, Haven, CT2020Google Wang al.43Wang Xue Overcapacity comprehensive assessment driving factors.Sustain. Times. 1426https://doi.org/10.3390/su13031426Crossref (4) ScholarStrict prevent induce They mandate expensive antipollution technologies, encouraging retire.We performance “Climate Change” (EPI) University effectiveness policies. Countries strict effective tend lower intensity score.F9: growthyes% growthEMBER (2021)29EMBERGlobal Jakob,16Manych Dorband al.44Dorband I.I. Unraveling insights 147111860https://doi.org/10.1016/j.enpol.2020.111860Crossref (29) growth creates production acceleration affordable technologies Under competitive.Model set:For without data, 2019.For Myanmar, latest 2014, so 2010–2014.F10: stabilityyes−2.5 (weak) 2.5 (strong)World (2021)45World BankWorld Development Indicators.2021Google ScholarUnstable socio-economic uncertainty, especially supportive attractive.Since 2019.F11: rentyes% GDPWorld Blondeel al.46Blondeel Graaf T. Haesebrouck Moving beyond explaining alliance.Energy 59101304https://doi.org/10.1016/j.erss.2019.101304Crossref (34) ScholarThe attractiveness source discourage CFPPs.We rents show contribution mining activity country’s economy. High active revenue country. reserves alone having significant policy.Since 2019.F12: Natural gas priceyesUS$/mmBtuBP ScholaryesRoss,34Ross Scholvin,38Scholvin Fell Kaffine,47Fell Kaffine D.T. fall joint impacts emissions.Am. 90-116https://doi.org/10.1257/pol.20150321Crossref (75) Gray Bernell48Gray Bernell Tree-hugging utilities? politics unusual alliance passed Oregon’s transition law.Energy 59101288https://doi.org/10.1016/J.ERSS.2019.101288Crossref natural decreases Henry Hub uniformly countries.F13: outputyes% generationEMBER IRENA (2021)49IRENAStatistical Profile.2021Google ScholaryesDodd Nelson,50Dodd Nelson Trials tribulations responses change: insight through transformation Australian market.Aust. Manag. 44: 614-631https://doi.org/10.1177/0312896219874096Crossref Kaffine47Fell decrease utilization retirement.For 2010–2019 cover countries.F14: Share nuclear mixyes% reliance preference its emission-free status.Excluded analysisPlant ageyes–Global MonitoryesMo Trencher al.,51Trencher Rinscheid Duygan Truong Revisiting systems: perpetuation Japan.Energy 69101770https://doi.org/10.1016/j.erss.2020.101770Crossref (45) Dodd Nelson50Dodd Scholar––Energy targetyes–NDCnoGallagher al.,41Gallagher