Draft:Cyril Voyant

• Comment: Sources need culling; most of the sources are not independent of the subject. Wikipedia doesn't care about what a subject has to say about himself by citing his own work, what matters more is coverage by sources independent of the subject. ~Anachronist (talk) 19:00, 7 August 2025 (UTC)

---

Cyril Voyant (in-progress, working about secondary references)

Cyril Voyant (born October 6, 1977) is a French physicist and Research Director at the Mines Paris – PSL. His career spans medical physics, radiotherapy, and solar energy forecasting, emphasizing hybrid AI models and energy system rmanagement^[1][2]. He holds two doctoral degrees in distinct disciplines. His first is a Diplôme de Qualification en Physique Radiologique et Médicale (DQPRM), the French national professional doctorate qualifying medical physicists to practice in clinical radiotherapy, nuclear medicine, and diagnostic imaging, with a focus on radiation–matter interaction and patient treatment planning. His second doctorate is a Ph.D. in applied mathematics, energy, and meteorology, focusing on statistical modelling, time-series analysis, and renewable energy resource forecasting. The combination of clinically and research oriented Ph.D. has shaped his interdisciplinary approach to both healthcare technologies and energy systems.

Clinical physics background

From 2003 to 2024, he served as a medical physicist (Diplôme de Qualification en Physique Radiologique et Médicale, DQPRM) in French hospitals, including leadership of the Medical Physics and Radiation Protection Unit at CHD Castelluccio in Ajaccio^[3]. The DQPRM is the French national professional doctorate granting legal qualification to practice in clinical radiotherapy, nuclear medicine, and diagnostic imaging, with a strong focus on patient safety, dosimetry, and radiation protection. He co-developed the open-source free-software LQL-Equiv (used in more than 20 countries), which computes biologically equivalent doses (BED/EQD2) using Linear-Quadratic-Linear (LQL) formalism, with repopulation correction and flexible fractionation^[4]. The LQL-Equiv model, as published in Clinical Oncology^[5], follows the BED/LQL formalism and introduces a simple optimisation on a cost function between regimens. Cost function between two regimens (1) and (2):

${\displaystyle f(n_{1},d_{1},ja_{1};\,n_{2},d_{2},ja_{2})\;=\;{\big |}\;\mathrm {BED} _{1}(n_{1},d_{1},ja_{1})\;-\;\mathrm {BED} _{2}(n_{2},d_{2},ja_{2})\;{\big |}}$

Equivalent-dose optimisation (reference at 2 Gy/fx named ${\displaystyle \mathrm {EQD} _{2}}$): find the number of 2 Gy fractions ${\displaystyle n_{0}}$ that minimises the cost: ${\displaystyle \operatorname {arg\,min} _{\,n_{\mathrm {ref} }\in \mathbb {R} ^{\ast }}\;f\!{\big (}n_{\mathrm {ref} },\,2;\,n,\,d,\,ja{\big )}\;=\;n_{0},\qquad \mathrm {EQD} _{2}\;=\;2\,n_{0}\;.}$

Here, ${\displaystyle n}$ is the number of fractions, ${\displaystyle d}$ the dose per fraction, ${\displaystyle \alpha /\beta }$ the tissue parameter, and ${\displaystyle ja}$ encodes days of interruption. ^[6] The tool has been adopted in numerous radiation therapy units worldwide, as evidenced by independent applications such as RTtxGap^[7].

He also contributed to modelling and potentially use the deposited dose from atmospheric neutrons (in flight) in the context of Boron Neutron Capture Therapy (BNCT) for therapeutic purposes. Done using atmospheric neutron flux spectra and nuclear cross-section data from the Los Alamos National Laboratory Evaluated Nuclear Data Files (ENDF/B-VI) and IAEA to model ${\displaystyle {^{10}\mathrm {B} }(n,\alpha ){^{7}\mathrm {Li} }}$ capture probabilities and other reactions ^[8][9]. He has also worked on the quantification of dose uncertainties in radiotherapy^[10].

Solar forecasting and energy research

His research focuses on probabilistic and AI-driven forecasting of solar and renewable energy resources for smart energy systems^[11]. The work combines time-series analysis, stochastic modelling, and machine learning with a strong emphasis on operational robustness.

Stochastic variability and forecastability. He introduced the Stochastic Coefficient of Variation (sCV) and a Forecastability Index (F) to quantify, respectively, variability and predictability of solar irradiance series using a clear-sky upper bound and autocorrelation structure^[12]. In that framework, with measured GHI ${\displaystyle I(t)}$ and clear-sky envelope ${\displaystyle I_{\mathrm {clr} }(t)}$, the variability is normalised as:

${\displaystyle \mathrm {sCV} \;=\;{\frac {\sqrt {\mathbb {E} {\big [}(\,I(t)-I_{\mathrm {clr} }(t)\,)^{2}{\big ]}}}{\displaystyle {\frac {|1-d|}{\sqrt {2}}}\;{\sqrt {\sigma _{\mathrm {clr} }^{2}+\mu _{\mathrm {clr} }^{2}}}}}\,,\quad d\in [0,1)\,,}$

where ${\displaystyle \mathbb {E} [\cdot ]}$ denotes the statistical expectation (mean), ${\displaystyle d}$ is the ratio between the minimum and maximum of the clear-sky signal over the considered cyclostationary period, ${\displaystyle \sigma _{\mathrm {clr} }}$ is the standard deviation of the clear-sky series over that period and ${\displaystyle \mu _{\mathrm {clr} }}$ is the mean of the clear-sky series over that period.

The numerator represents the root mean square (RMS) of the deviation between the actual and clear-sky signals, which reflects residual variability due to clouds or atmospheric instability. The denominator scales this RMS by the variability and mean level of the clear-sky signal, providing a dimensionless metric that can be compared across sites and time scales.

The forecastability uses the residual autocorrelation (maximum over lags within a cyclostationary period): ${\displaystyle F\;=\;(1-\mathrm {sCV} )\;+\;\rho _{\max }\,\mathrm {sCV} \,.}$

ClearSky-Free ELMs and cyclostationarity. He proposed a ClearSky-Free approach using Extreme Learning Machines trained directly on raw irradiance, where multiple seasonal/diurnal cycles are embedded as features (cyclostationary design), removing the need for prior stationarisation or clear-sky normalisation^[13]. Training uses a Ridge-regularised objective (least squares + L2):

${\displaystyle J(\theta )\;=\;{\frac {1}{N}}\sum _{t=1}^{N}{\big (}y_{t}-{\hat {y}}_{t}{\big )}^{2}\;+\;\lambda \,\|\theta \|_{2}^{2}\,,}$

where ${\displaystyle N}$ is the number of training samples, ${\displaystyle y_{t}}$ and ${\displaystyle {\hat {y}}_{t}}$ are respectively the observed and predicted values at time ${\displaystyle t}$, ${\displaystyle \lambda }$ is the regularisation coefficient, and ${\displaystyle \|\theta \|_{2}^{2}}$ is the squared L2-norm of the model parameters. A non-parametric lookup-table procedure maps predictions to empirical predictive intervals, avoiding distributional assumptions.

Error analysis (bias–variance decomposition). To diagnose model behaviour (under/overfitting vs intrinsic noise) in solar resource forecasting, he formalised the mean squared error decomposition: ${\displaystyle \mathrm {MSE} \;=\;{\big (}\mathrm {Bias} [\,{\hat {y}}\,]{\big )}^{2}\;+\;\mathrm {Var} [\,{\hat {y}}\,]\;+\;\sigma _{\varepsilon }^{2}\,,}$ and applied it systematically to benchmarking exercises across climates and horizons^[14]. This allows separating reducible error (bias/variance) from irreducible measurement/process noise, with direct implications for model selection and horizon-dependent tuning.

Transfer learning and spatial adaptation. He used transfer learning and spatial clustering to adapt models to data-sparse regions, exploiting geographical similarity for efficient fine-tuning while preserving generalisation^[15].

Complex-valued forecasting. He introduced a compact complex representation where the real part encodes the (normalised) irradiance signal and the imaginary part encodes a data-driven volatility measure computed on a sliding window. A linear complex autoregression then provides joint point/probabilistic forecasts: ${\displaystyle z(t)\;=\;\kappa (t)\;+\;j\,\sigma _{\tau }(t)\,,\qquad z_{b}(t\!+\!1)\;=\;\sum _{i=0}^{p-1}\omega _{i}\,z(t-i)\,,}$ with complex weights ${\displaystyle \omega _{i}}$, clear-sky index ${\displaystyle \kappa (t)}$ and local volatility ${\displaystyle \sigma _{\tau }(t)}$^[16].

In a 2014 Applied Energy study, Voyant et al. used satellite-based ANN models (~3km x 3km) to predict hourly solar radiation in Corsica with ≈16.5 % nRMSE.^[17]

Overall, this part presents a sample of contributions (from non-parametric variability/forecastability metrics and cyclostationary ELMs to rigorous error decomposition and complex-domain forecasting) aimed at mathematically grounded, in solar prediction context.

Foundational machine-learning contributions

Machine learning has become an essential tool in solar radiation forecasting since the early 2000s. Significant contributions in this domain have been made partly by Philippe Lauret (Université de la Réunion), who pioneered the use of neural networks for cyclostationary solar signals. A benchmarking study co-authored with P. Lauret (Solar Energy, 2015) has over 298 citations, reflecting its impact on ML-based solar forecasting.^[18]. Another pionners in topic are Philippe Blanc (Mines Paris – PSL) on hybrid physical–statistical forecasting models^[19], Richard Perez (University at Albany) who advanced transposition and diffuse sky models, which are foundational in solar irradiance modeling and nowcasting techniques^[20], and Dazhi Yang (Harbin Institute of Technology), whose influential works include a review of solar forecasting connecting atmospheric science perspectives with grid-integration needs^[21], as well as methodological benchmarks in hourly ML-based forecasting across multiple climates^[22].

In this context, Cyril Voyant co-authored one of the most cited reviews (~2000 citations) on machine-learning methods for solar radiation forecasting^[23]. His research systematically explores a broad range of techniques: from neural networks and extreme learning machines to support vector regression, random forests, ensemble learning, and hybrid physical–statistical approaches^[24]. Further contributions include the introduction of forecastability metrics, bias–variance–noise decomposition, transfer learning between sites, cyclostationary modelling, and complex-valued time series. Collectively, these studies form part of a coherent and collaborative effort, alongside Lauret, Blanc, Perez, Yang and others, to enhance the accuracy, robustness, and interpretability of renewable energy forecasting.

Collaborative projects^[25]

He contributed to several significant research initiatives like SAPHIR (ANR)^[26], Fine4cast (PEPR TASE)^[27] and TILOS (Horizon 2020 EU)^[28].

Recognition

Cyril Voyant appears in the list of Stanford University Top 2 % Most‑Cited Scientists globally in the field of Energy/Engineering since 2020, according to the list maintained by Elsevier and Stanford University^[29][30][31] and in the top 20 solar forecasting researchers ranked based on number of appearances in the first 1000 results returned by Google Scholar^[32]. These distinctions, often used as a marker of significant research impact in energy forecasting. Indeed, as of 2025, independent citation databases list over 7,100 citations to his publications and an h-index above 30^[33][34].

One of his papers was selected among the Top 100 most downloaded cancer research articles in Scientific Reports by Nature in 2024.^[35]

He serves on editorial boards, including the Journal of Radiology and Oncology^[36] and has participated in scientific committees for some international conferences (IWCMC, IWCF, ENVIROSENS, etc.).

He received the City of Nice Prize in 2012 from the Accademia Corsa for his doctoral thesis on solar irradiance time-series prediction using artificial neural networks^[37]. Nice-Matin, a major French newspaper, reported that he received this prize for his PhD thesis on “solar irradiance and PV output forecasting with neural networks"^[38]. A competitive award for young researchers signaling the significance of his early work in renewable energy forecasting.

He was also recognized at the PVSEC conference European Photovoltaic Solar Energy Conference (Student Award at the 25th EU PVSEC 2010)^[39].

References

1. "ORCID". orcid.org. Retrieved 2025-08-07.
2. "Voyant, Cyril". www.idref.fr. Archived from the original on 2024-10-05. Retrieved 2025-08-07.
3. "Cyril Voyant". www.cyrilvoyant.com. Retrieved 2025-08-07.
4. Voyant, Cyril; Julian, Daniel (2025-08-04), LQL-Equiv: Open-Source Software for Biologically Equivalent Dose Calculation in Radiotherapy, doi:10.5281/zenodo.16739882, retrieved 2025-08-07
5. Voyant, Cyril; Julian, Daniel (2025). "Improving Clinical Decision-Making in Radiotherapy: A Comparative Analysis of LQ and LQL Dose Models". Clinical Oncology. doi:10.1016/j.clon.2025.103893. PMID 40680517.
6. Voyant, C.; Julian, D. (2017-04-15). "A Short Synthesis Concerning Biological Effects and Equivalent Doses in Radiotherapy". arXiv:1704.06457 [physics.med-ph].
7. Yusoff, A L; Mohamad, M; Abdullah, R; Bhavaraju, V M K; Idris, N R Nik (March 2016). "RTtxGap: An android radiobiological tool for compensation of radiotherapy treatment interruption". Journal of Physics: Conference Series. 694 (1): 012012. Bibcode:2016JPhCS.694a2012Y. doi:10.1088/1742-6596/694/1/012012. ISSN 1742-6588.
8. Voyant, Cyril; Roustit, Rudy; Tatje, Jennifer; Biffi, Katia; Leschi, Delphine; Briançon, Jérome; Marcovici, Céline Lantieri (January 2011). "Therapeutic potential of atmospheric neutrons". Reports of Practical Oncology & Radiotherapy. 16 (1): 21–31. doi:10.1016/j.rpor.2010.11.002. PMC 3920293. PMID 24669300.
9. "ENDF/B-VI Nuclear Data". Los Alamos National Laboratory. Retrieved 2025-08-09.
10. Voyant, C.; Biffi, K.; Leschi, D.; Briançon, J.; Lantieri, C. (2011-07-01). "Dosimetric uncertainties related to the elasticity of bladder and rectal walls: Adenocarcinoma of the prostate". Cancer/Radiothérapie. 15 (4): 270–278. arXiv:1104.0782. doi:10.1016/j.canrad.2010.12.006. ISSN 1278-3218. PMID 21497533.
11. "Google Scholar". scholar.google.com. Retrieved 2025-08-07.
12. Voyant, Cyril; Julien, Alan; Despotovic, Milan; Notton, Gilles; Garcia-Gutierrez, Luis Antonio; Nicolosi, Claudio Francesco; Blanc, Philippe; Bright, Jamie (January 2026). "Stochastic coefficient of variation: Assessing the variability and forecastability of solar irradiance". Renewable Energy. 256 123913. Bibcode:2026REne..25623913V. doi:10.1016/j.renene.2025.123913.
13. Voyant, Cyril; Despotovic, Milan; Notton, Gilles; Saint-Drenan, Yves-Marie; Asloune, Mohammed; Garcia-Gutierrez, Luis (July 2025). "On the importance of clearsky model in short-term solar radiation forecasting". Solar Energy. 294 113490. arXiv:2503.07647. Bibcode:2025SoEn..29413490V. doi:10.1016/j.solener.2025.113490.
14. Voyant, Cyril; Notton, Gilles; Duchaud, Jean-Laurent; Lauret, Philippe; Garcia-Gutierrez, Luis; Faggianelli, Ghjuvan Antone (2022). "Benchmarks for solar radiation time series forecasting". Renewable Energy. 191: 747–762. arXiv:2203.14959. Bibcode:2022REne..191..747V. doi:10.1016/j.renene.2022.04.065.
15. Despotovic, Milan; Voyant, Cyril; Garcia-Gutierrez, Luis; Almorox, Javier; Notton, Gilles (July 2024). "Solar irradiance time series forecasting using auto-regressive and extreme learning methods: Influence of transfer learning and clustering". Applied Energy. 365 123215. Bibcode:2024ApEn..36523215D. doi:10.1016/j.apenergy.2024.123215.
16. Voyant, Cyril; Lauret, Philippe; Notton, Gilles; Duchaud, Jean-Laurent; Garcia-Gutierrez, Luis; Faggianelli, Ghjuvan Antone (2022-11-01). "Complex-valued time series based solar irradiance forecast". Journal of Renewable and Sustainable Energy. 14 (6) 066502. doi:10.1063/5.0128131.
17. Voyant; Haurant (2014). "Time series modeling and large scale global solar radiation forecasting from geostationary satellites data". Solar Energy. 102: 131. arXiv:1401.4644. Bibcode:2014SoEn..102..131V. doi:10.1016/j.solener.2014.01.017.
18. Lauret, Philippe; Voyant, Cyril; Soubdhan, Ted; David, Mathieu; Poggi, Philippe (February 2015). "Lauret et al (2015) – Solar Energy citations". Solar Energy. 112: 446–457. Bibcode:2015SoEn..112..446L. doi:10.1016/j.solener.2014.12.014. Retrieved 2025-08-09.
19. Lauret, Philippe; Voyant, Cyril; Soubdhan, Ted; David, Mathieu; Poggi, Philippe (2015). "A benchmarking of machine learning techniques for solar radiation forecasting in an insular context". Solar Energy. 112: 446–457. Bibcode:2015SoEn..112..446L. doi:10.1016/j.solener.2014.12.014.
20. Lim, Su Pei; Pandikumar, Alagarsamy; Lim, Hong Ngee; Huang, Nay Ming (2016). "Essential role of N and Au on TiO2 as photoanode for efficient dye-sensitized solar cells". Solar Energy. 125: 135–145. Bibcode:2016SoEn..125..135L. doi:10.1016/j.solener.2015.12.019.
21. Yang, Dazhi; Wang, Wenting; Gueymard, Christian A.; Hong, Tao; Kleissl, Jan; Huang, Jing; Perez, Marc J.; Perez, Richard; Bright, Jamie M.; Xia, Xiang'ao; Van Der Meer, Dennis; Peters, Ian Marius (2022). "A review of solar forecasting, its dependence on atmospheric sciences and implications for grid integration: Towards carbon neutrality". Renewable and Sustainable Energy Reviews. 161 112348. Bibcode:2022RSERv.16112348Y. doi:10.1016/j.rser.2022.112348.
22. Yagli, Gokhan Mert; Yang, Dazhi; Srinivasan, Dipti (2019). "Automatic hourly solar forecasting using machine learning models". Renewable and Sustainable Energy Reviews. 105: 487–498. Bibcode:2019RSERv.105..487Y. doi:10.1016/j.rser.2019.02.006.
23. Voyant, Cyril; Notton, Gilles; Kalogirou, Soteris; Nivet, Marie-Laure; Paoli, Christophe; Motte, Fabrice; Fouilloy, Alexis (2017). "Machine learning methods for solar radiation forecasting: A review". Renewable Energy. 105: 569–582. Bibcode:2017REne..105..569V. doi:10.1016/j.renene.2016.12.095.
24. "Archive ouverte HAL". cv.hal.science. Retrieved 2025-08-09.
25. Миљојковић, Јасмина. "Faculty of Engineering Strengthens International Research Collaboration and Excellence in Research". www.fin.kg.ac.rs. Retrieved 2025-08-11.
26. "Prévisions météorologiques a haute-résolution améliorée par des données capteurs". Agence nationale de la recherche (in French). Retrieved 2025-08-07.
27. "FINE4CAST". PEPR TASE. Retrieved 2025-08-07.
28. "Technology Innovation for the Local Scale, Optimum Integration of Battery Energy Storage | TILOS | Projet | Résultats | H2020 | CORDIS | Commission européenne". CORDIS | European Commission (in French). Retrieved 2025-08-07.
29. List, Top Scientists. "World's Top 2% Scientists". topresearcherslist.com. Retrieved 2025-08-07.
30. rédaction, La. "Quatre chercheurs de l'Université de Corse parmi les 2% les plus influents au monde". Corse Net Infos - Pure player corse (in French). Retrieved 2025-08-07.
31. "Université : trois chercheurs distingués à l'international". www.corsematin.com (in French). 2021-12-05. Retrieved 2025-08-07.
32. Yang, Dazhi; Kleissl, Jan; Gueymard, Christian A.; Pedro, Hugo T.C.; Coimbra, Carlos F.M. (July 2018). "History and trends in solar irradiance and PV power forecasting: A preliminary assessment and review using text mining". Solar Energy. 168: 60–101. Bibcode:2018SoEn..168...60Y. doi:10.1016/j.solener.2017.11.023.
33. Index, AD Scientific. "Cyril Voyant – Others | Université de Corse Pasquale Paoli". AD Scientific Index. Retrieved 2025-08-09.
34. "Cyril Voyant". www.wikidata.org. Retrieved 2025-08-09.
35. "Cancer Top 100 of 2024 – Nature". Nature. 5 March 2025. Retrieved 2025-08-07.
36. "Cyril Voyant | Cyril Voyant, University of Corsica | Journal of Radiology and Oncology | HSPI". www.radiooncologyjournal.com. Retrieved 2025-08-07.
37. rédaction, La (2012-06-27). "L'Accademia corsa de Nice a remis ses prix 2012". Nice-Matin (in French). Retrieved 2025-08-07.
38. "Cyril Voyant". theses.fr (in French). Retrieved 2025-08-09.
39. "PVSEC | Awards". PVSEC (in Japanese). Retrieved 2025-08-07.