ENR et réseaux de neurones

Time series modeling and large scale global solar radiation forecasting from geostationary satellites data.

Cyril Voyant, Pierrick Haurant, Marc Muselli, Christophe Paoli, Marie-Laure Nivet


Nouvel article publié dans la revue Solar Energy
Dans ce nouveau papier, nous avons développé une nouvelle méthode de prévision du rayonnement global sur une large échelle (toute la Corse). Le principe repose sur l'utilisation conjointe des outils stochastiques de prédictions des séries temporelles et l'utilisation des images satellites.
Avec cette approche, chaque pixel (Cf base de données Helioclim 3; 1158 pixels par images horaire) devient support d'une série temporelle, ce qui implique que pour une prévision 1 heure à l'avance, le processus doit générer plus de 1100 prédictions 1D. Nous avons montré que cette approche est compétitive à l'issu d'un test effectué sur deux années complètes de rayonnement global mesuré.
La prochaine étape sera de confronter cette méthodologie de prédiction avec des plus globales comme peuvent l'être les NWP, nous comparerons ainsi notre approche avec les modèles AROME et NOAA.
L'abstract du manuscrit est:
"When a territory is poorly instrumented, geostationary satellites data can be useful to predict global solar radiation. In this paper, we use geostationary satellites data to generate 2-D time series of solar radiation for the next hour. The results presented in this paper relate to a particular territory, the Corsica Island, but as data used are available for the entire surface of the globe, our method can be easily exploited to another place. Indeed 2-D hourly time series are extracted from the HelioClim-3 surface solar irradiation database treated by the Heliosat-2 model. Each point of the map have been used as training data and inputs of artificial neural networks (ANN) and as inputs for two persistence models (scaled or not). Comparisons between these models and clear sky estimations were proceeded to evaluate the performances. We found a normalized root mean square error (nRMSE) close to 16.5% for the two best predictors (scaled persistence and ANN) equivalent to 35-45% related to ground measurements. Finally in order to validate our 2-D predictions maps, we introduce a new error metric called the gamma index which is a criterion for comparing data from two matrixes in medical physics. As first results, we found that in winter and spring, scaled persistence gives the best results (gamma index test passing rate is respectively 67.7% and 86%), in autumn simple persistence is the best predictor (95.3%) and ANN is the best in summer (99.8%)."

https://www.researchgate.net/publication/259759274_Time_series_modeling_and_large_scale_global_solar_radiation_forecasting_from_geostationary_satellites_data?ev=prf_pub https://www.researchgate.net/publication/259759274_Time_series_modeling_and_large_scale_global_solar_radiation_forecasting_from_geostationary_satellites_data?ev=prf_pub



Samedi 18 Janvier 2014

Bayesian rules and stochastic models for high accuracy prediction of solar radiation
C Voyant, C Darras, M Muselli, C Paoli, ML Nivet and P Poggi


Nouvel article accepté dans la revue "Applied Energy"
Dans cet article, nous avons tenté de prédire la ressource solaire 24 heures à l'avance en utilisant les réseaux de neurones, les processus ARMA et les inférences Bayésiennes.
Cet outil prédictif pourrait à terme être repris dans le code de calcul ORIENTE dédié au pilotage de la station MYRTHE (PV sur Ajaccio). Il est en effet impératif de connaitre ce que seront les réserves énergétiques (pile à combustible Vs électrolyseur) pour bien quantifier la part qui pourra être envoyée sur le réseau.

Le résumé de cet article est :
"It is essential to find solar predictive methods to massively insert renewable energies on the electrical distribution grid. The goal of this study is to find the best methodology allowing predicting with high accuracy the hourly global radiation. The knowledge of this quantity is essential for the grid manager or the private PV producer in order to anticipate fluctuations related to clouds occurrences and to stabilize the injected PV power. In this paper, we test both methodologies: single and hybrid predictors. In the first class, we include the multi-layer perceptron (MLP), auto-regressive and moving average (ARMA), and persistence models. In the second class, we mix these predictors with Bayesian rules to obtain ad-hoc models selections, and Bayesian averages of outputs related to single models. If MLP and ARMA are equivalent (nRMSE close to 40.5% for the both), this hybridization allows a nRMSE gain upper than 14 percentage points compared to the persistence estimation (nRMSE=37% versus 51%)."

manuscript.pdf manuscript.pdf  (888.34 Ko)
https://www.researchgate.net/publication/256663517_Bayesian_rules_and_stochastic_models_for_high_accuracy_prediction_of_solar_radiation?fulltextDialog=true https://www.researchgate.net/publication/256663517_Bayesian_rules_and_stochastic_models_for_high_accuracy_prediction_of_solar_radiation?fulltextDialog=true


The first International Conference on Nanoelectronics, Communications and Renewable Energy (ICNCRE’13) will be held on September 22-23 2013 at the house of culture, Jijel, Algeria


La conférence de Jijel sera l'occasion de montrer les résultats d'une étude sur la sélection de modèles prédictifs basés sur l'utilisation de l'information mutuelle. Le résumé de cet article est:

"Numerous methods exist and were developed for global radiation forecasting. The two most popular types are the numerical weather predictions (NWP) and the predictions using stochastic approaches. This article presents a methodology for determining the best method to use, according to a rule related to the spatial resolution, temporal step and location. We propose to compute a parameter noted  constructed in part from the mutual information which is a quantity that measures the mutual dependence of two variables. Both of these are calculated with the objective to establish the more relevant method between NWP and stochastic models concerning the current problem."

Outre l'aspect sélection, nous montrerons aussi une nouvelle méthodologie d'inclinaison du rayonnement global utilisant les ANN, ci-dessous le résumé:

"Calculation of solar global irradiation on tilted planes from horizontal global one is difficult when the time step is small.We used an Artificial Neural Network (ANN) to realize this conversion at a 5-min time step for solar irradiation data of Bouzareah (Algeria). The ANN is developed and optimized on the basis of two years of solar data (1.5 year or training and 0.5 year for test) and the accuracy of the optimal configuration is around 8% for the nRMSE."

https://www.researchgate.net/publication/253234194_Estimation_of_5-min_solar_global_irradiation_on_tilted_planes_by_ANN_method_in_Bouzareah_Algeria?ev=prf_pub https://www.researchgate.net/publication/253234194_Estimation_of_5-min_solar_global_irradiation_on_tilted_planes_by_ANN_method_in_Bouzareah_Algeria?ev=prf_pub
https://www.researchgate.net/publication/253233999_The_global_radiation_forecasting_based_on_NWP_or_stochastic_modeling_an_objective_criterion_of_choice?ev=prf_pub https://www.researchgate.net/publication/253233999_The_global_radiation_forecasting_based_on_NWP_or_stochastic_modeling_an_objective_criterion_of_choice?ev=prf_pub



Lundi 29 Juillet 2013

Multi-horizon solar radiation forecasting for Mediterranean locations using time series models
C. Voyant, C. Paoli, M. Muselli and ML. Nivet


Article accepté dans "Renewable and Sustainable Energy Reviews"
Cet article synthétise les résultats de prédiction de rayonnement solaire que l'on a pu avoir ces dernières années. Tous les horizons de prédiction que l'on a étudiés (de 5 minutes à 24 heures à l'avance) sont classés sous forme d'une review générale. Le résumé de cette publication est :

Considering the grid manager's point of view, needs in terms of prediction of intermittent energy like the photovoltaic resource can be distinguished according to the considered horizon: following days (d+1, d+2 and d+3), next day by hourly step (h+24), next hour (h+1) and next few minutes (m+5 e.g.). Through this work, we have identified methodologies using time series models for the prediction horizon of global radiation and photovoltaic power. What we present here is a comparison of different predictors developed and tested to propose a hierarchy. For horizons d+1 and h+1, without advanced ad hoc time series pre-processing (stationarity) we find it is not easy to differentiate between autoregressive moving average (ARMA) and multilayer perceptron (MLP). However we observed that using exogenous variables improves significantly the results for MLP . We have shown that the MLP were more adapted for horizons h+24 and m+5. In summary, our results are complementary and improve the existing prediction techniques with innovative tools: stationarity, numerical weather prediction combination, MLP and ARMA hybridization, multivariate analysis, time index, etc.

https://www.researchgate.net/publication/250612118_Multi-horizon_solar_radiation_forecasting_for_Mediterranean_locations_using_time_series_models https://www.researchgate.net/publication/250612118_Multi-horizon_solar_radiation_forecasting_for_Mediterranean_locations_using_time_series_models



Dimanche 21 Juillet 2013

Clear sky models assessment for an operational PV production forecasting solution

Sylvain Cros , Olivier Liandrat, Nicolas Sébastien, Cyril Voyant, Nicolas Schmutz


Second poster accepté à la conférence PVSEC de Villepinte 2013
Ce sera donc le second papier auquel j'aurai contribué qui sera proposé lors de cette conférence internationale. Un grand merci à l'équipe Reuniwatt qui est à la base de ce projet de comparaison de modèles de ciel clair. Le résumer de cette présentation est :

Photovoltaic power production relies upon solar radiation received at ground level, that remains
difficult to predict. Accurate forecast of surface solar irradiance is essential for grid operators in
order to accommodate this intermittent energy in their scheduling, dispatching and regulation of
power. Surface solar irradiance forecasting methods can be described as the prediction of the cloud
physical property over a specific area. This information is then combined with the value of the
irradiance under a clear sky for the same area at the same forecast time. Then, uncertainty of
irradiance under clear sky can affect significantly the forecast results accuracy.
Many clear sky models have been designed and are routinely used to compute surface solar
irradiance under a sky with no cloud for a diversity of applications. The required inputs for such
tools are varying from model to another. The simplest models just take into account the solar elevation, while more detailed ones may include other inputs such as concentration of atmospheric
components (aerosols, water vapor, ozone), elevation of the site, ground albedo, in order to
represent a realistic atmosphere transmittance. However, availability and quality of some inputs are
not always guaranteed at every locations.
The objective of our study is the selection and performance quantification of a clear sky model
usable with an innovative solar radiation forecasting tool. We selected the models Bird, Solis, ESRA
and McClear. These models demonstrated their accuracy and robustness and their features are
suitable for operational processes. We undertook our study over three sites located in Corsica,
French Guiana and Reunion Island. We collected available atmospheric parameters (aerosol optical
depth and water vapor) at these locations through the AERONET network and the MACC-IFS
reanalysis outputs. We compared simulated global irradiation with ground measurements at clear
sky situations. Comparisons have been made at a time step of one minute.
In our results, we present the intrinsic accuracy of each models for the three different sites. We
demonstrate explicitly that atmospheric inputs data quality are more important than the choice of
the clear sky model itself. However, for each models we quantified the loss of accuracy for clear
sky radiation according to the time sampling of atmospheric input data (monthly and daily). Finally,
we present recommendations for an efficient clear sky model setting according to the effective
availability of atmospheric input parameters at a given location.

5bv_4_69.pdf 5BV.4.69.pdf  (258.63 Ko)



Dimanche 16 Juin 2013

Comparison of clear-sky models performance for a photovoltaic energy production forecasting tool

S. Cros, O. Liandrat, C. Voyant, and N. Schmutz


Les membres de l'équipe Reuniwatt auront l'occasion de montrer une comparaison de modèles de rayonnement global par ciel clair lors du 13th European Meteorological Society (EMS) annual Meeting & 11th European Conference on Applications of Meteorology (ECAM). Je suis heureux d'avoir pu participer à cette étude et surtout de pouvoir intégrer une conférence de ce type. Voici le résumé de la présentation:

"Photovoltaic power production relies upon solar radiation received at ground level, that remains difficult to predict. Accurate forecast of surface solar irradiance is essential for grid operators in order to accommodate this intermittent energy in their scheduling, dispatching and regulation of power. Surface solar irradiance forecasting methods can be described as the prediction of the cloud physical property over a specific area. This information is then combined with the value of the irradiance under a clear sky for the same area at the same forecast time. Then, uncertainty of irradiance under clear sky can affect significantly the forecast results accuracy.
Many clear sky models have been designed and are routinely used to compute surface solar irradiance under a sky with no cloud for a diversity of applications. The required inputs for such tools are varying from model to another. The simplest models just take into account the solar elevation, while more detailed ones may include other inputs such as concentration of atmospheric components (aerosols, water vapor, ozone), elevation of the site, ground albedo, in order to represent a realistic atmosphere transmittance. However, availability and quality of some inputs are not always guaranteed at every locations.
The objective of our study is the selection and performance quantification of a clear sky model usable with an innovative solar radiation forecasting tool. We selected the models Bird, Solis, ESRA and McClear. These models demonstrated their accuracy and robustness and their features are suitable for operational processes. We undertook our study over three sites located in Corsica, French Guiana and Reunion Island. We collected available atmospheric parameters (aerosol optical depth and water vapour). We compared simulated global irradiation with ground measurements at clear sky situations. Models accuracy and their sensitivity to input data are discussed. Conclusions underline the input data quality requirements."


Mardi 7 Mai 2013

URBAN OZONE CONCENTRATION FORECASTING WITH ARTIFICIAL NEURAL NETWORK IN CORSICA

WANI TAMAS, GILLES NOTTON, CHRISTOPHE PAOLI, MARIE-LAURE NIVET, AURELIA BALU, CYRIL VOYANT


Participation à l'EENVIRO'13 qui aura lieu à Bucarest (Roumanie) en Septembre 2013
Cette présentation sera l'occasion pour Wani Tamas (étudiant en thèse à l'UDC) de montrer l'intérêt des MLP et de l'analyse en composante principale pour la prédiction de séries temporelles de concentration de polluants atmosphériques. Le résumé de cet article est:

"Atmospheric pollutants concentration forecasting is an important issue in air quality monitoring. Qualitair Corse, the organization responsible for monitoring air quality in Corsica (France) region , needs to develop a short-term prediction model to lead its mission of information towards the public. Various deterministic models exist for meso-scale or local forecasting, but need powerful large variable sets, a good knowledge of atmospheric processes, and can be inaccurate because of local climatical or geographical particularities, as observed in Corsica, a mountainous island located in a Mediterranean Sea. As a result, we focus in this study on statistical models, and particularly Artificial Neural Networks (ANN) that have shown good results in the prediction of ozone concentration at horizon h+1 with data measured locally. The purpose of this study is to build a predictor to realize predictions of ozone and PM10 at horizon d+1 in Corsica in order to be able to anticipate pollution peak formation and to take appropriated prevention measures. Specific meteorological conditions are known to lead to particular pollution event in Corsica (e.g. Saharan dust event). Therefore, several ANN models will be used, for meteorological conditions clustering and for operational forecasting."

eenviro_2013.pdf EENVIRO 2013.pdf  (351.52 Ko)



Mardi 7 Mai 2013

Time series modeling with pruned multi-layer perceptron and 2-stage damped least-squares method

Voyant C, Tamas W, Paoli C, Muselli M, Nivet M, Notton G


Participation à une conférence de Mathématiques (appliquées à la physique) : l'IC-Msquare 2013 de Prague

final_paper.doc final_paper.doc  (276.5 Ko)

Cette conférence ("International Conference on Mathematical Modeling in Physical Sciences) sera à l'occasion pour nous de présenter nos résultats concernant un nouvel algorithme d'apprentissage des MLP.
Ce dernier baptisé "2-stage damped least-squares method" permet d'obtenir des réseaux "prunés" automatiquement.
Cet outil permet de faciliter le temps d'optimisation et d'améliorer le recherche des valeurs de poids et de biais lors de modélisation de séries temporelles météorologiques et de prédiction à horizon. L'abstract de cette présentation est:

"A Multi-Layer Perceptron (MLP) defines a family of artificial neural networks often used in TS modeling. Because of the “black boxes” aspect, many researchers refuse to use it. Moreover, the optimization (often based on the exhaustive approach where “all” configurations are tested) and learning phases of this artificial intelligence tool (often based on the Levenberg-Marquardt algorithm; LMA) are weaknesses of this approach. These two tasks must be repeated depending on the knowledge of each new problem studied, making the process, long, laborious and not robust. In this paper a pruning process is proposed. This method allows, during the training phase, to carry out an inputs selecting method activating (or not) inter-nodes connections and to verify both if optimization phase is shorten and forecasting are improved. We propose to use iteratively the popular damped least-squares method to activate inputs and neurons. A first pass is applied to 10% of the learning sample to determine weights significantly different from 0 and delete other. Then a classical batch process based on LMA is used with the new MLP. We validated our method on 10 measured meteorological TS with different behaviors (stationarity and periodicity). We used 8 years for the training and 2 years for the test, results of the classical LMA and the 2-stage LMA are cross-compared."

Hourly global radiation prediction from geostationary satellite data
P. Haurant, C. Voyant, M. Muselli, C. Paoli, ML Nivet


"Visual presentation" acceptée lors l'EU-PVSEC 2013 à Paris
Nous aurons l'occasion de montrer lors de cette conférence la nouvelle méthodologie de prédiction du rayonnement global que nous avons développée. Cette dernière reprend l'approche 1-d basée sur la modélisation stochastique des séries temporelles de mesures au sol. Cette fois, il n'est plus question de détecteurs, mais d'estimation d'irradiance via la database Helioclim-3. Chaque pixel de l'image (1158 points qui quadrille la corse) devient un générateur de séries temporelles. Les PMC, ARMA et autres modèles statistiques deviennent de ce fait applicables à ce nouveau problème 2-d et permettent de générée des cartes de rayonnement global à l'horizon 1 heure.
Voici l'abstract du papier :

"Intermittency and cyclical nature of global radiation at ground level limit the photovoltaic (PV) systems integration on electrical grids. Indeed, voltage instability is considered as one of the main threats for power systems security around the world. To overcome this problem, irradiation prediction is essential to drive the transition between renewable and non-renewable energy sources. Concerning the small electrical networks with a few PV stations, conventional tools using one measured time series per location allow radiation forecasting. This 1-D approach becomes obsolete for the large electrical networks with many stations. In this paper we generalize the 1-D time series methodology with the robustness of the satellite acquisitions. A new 2-D prediction tool combining HelioClim3 estimations and stochastic models is introduced and validated. With this study, the PV part of the global energy mix could be increased, reducing thereby the non-renewable and exhaustible fossil fuel part."

pv_sec_2013.pdf PV-SEC 2013.pdf  (12.13 Ko)



September 22-23, 2013, House of Culture, Jijel Algeria


Participation au comité scientifique d'une conférence internationale
L'université de Jijel organise une conférence internationale sur la thématique de la nanoélectronique et des énergies renouvelables. Concernant ce dernier topic, de nombreux experts viendront partager leur expérience en terme de modélisation et de prédiction de la ressource solaire. Cette étape est primordiale avant de pouvoir envisager une parfaite intégration de ce type d'énergie au sein des réseaux électriques.

24-hours ahead global irradiation forecasting using Multi-Layer Perceptron

Cyril Voyant, Prisca Randimbivololona, Marie-Laure Nivet, Christophe Paoli, Marc Muselli


Papier accepté dans la revue "meteorological applications" éditée par Wiley
Dans ce document, l'approche basée sur l’utilisation de réseaux de neurones pour prédire par pas horaire 24 heures à l'avance le rayonnement global est validée.
La méthodologie, bien que plus compliquée que dans le cas h+1 et d+1, donne des résultats qui égale voire améliore ceux que l'on peut trouver dans la littérature.
L'horizon considéré, est une clé de l'intégration de l'énergie PV au sein des réseaux électriques. En effet, dans le cas d'une station PV et d'une redistribution sur le réseau, il faut que l'exploitant soit capable le soir à 18h00 de prévoir ces courbes de charges pour la journée du lendemain. En cas de surestimation, des pénalités sont appliquées et en cas de sous-estimation, une partie de la puissance n'est pas envoyée, entrainant une perte immédiate.
Des études (en cours) vont permettre de tester cette méthodologie de prédiction sur le cas réel de station Myrte du laboratoire de Vignola sur Ajaccio.
Ci-dessous, l'abstract de la publication :


"The grid integration of variable renewable energy sources implies that their effective production could be predicted, at different times ahead. In the case of solar plants, the driving factor is the global solar irradiation (sum of direct and diffuse solar radiation projected on a plane (Wh/m²)). This paper focuses on the 24-hours ahead forecast of global solar irradiation (i.e. hourly solar irradiation prediction for the day after). A method based on artificial intelligence using Artificial Neural Network (ANN) is reported. The ANN hereafter considered is a Multi-Layer Perceptron (MLP) applied to a pre-treated time series (TS). Two architectures are tested; it is shown that the most relevant is based on a multi-output MLP using endogenous and exogenous input data. A real case 2-years TS is computed and the MLP results are compared with both a statistical approach (AutoRegressive-Moving Average model; ARMA) and a reference persistent approach. Results show that the prediction error estimate (nRMSE) can be reduced by 1.3 points with an ANN compared to ARMA and by 7.8 points compared to the naïve persistence."


https://www.researchgate.net/publication/234037098_24-hours_ahead_global_irradiation_forecasting_using_Multi-Layer_Perceptron?ev=prf_pub https://www.researchgate.net/publication/234037098_24-hours_ahead_global_irradiation_forecasting_using_Multi-Layer_Perceptron?ev=prf_pub



Dimanche 24 Février 2013

Multi-horizon irradiation forecasting for Mediterranean locations using time series models

Christophe Paoli, Cyril Voyant, Marc Muselli, Marie-Laure Nivet et Pierrick Haurant


Communication orale acceptée au prochain "Solar world congress 2013" proposé par l'International Energy Society à Cancun (Mexique)
Nous proposerons dans cette conférence une synthèse des différents travaux que nous avons déjà menés en ajoutant de nouvelles méthodologies de prédictions pour le très court terme (5 minutes à l'avance).
Un papier construit sous forme d'une "review" de tout ce que l'on a proposé dans des articles précédents, devrait à l'issu de cette conférence, voir le jour.
Il sera alors envisageable, une fois ces différentes étapes achevées, de considérer les prédictions non plus sur un site unique, mais un territoire complet en utilisant des images satellites et des interprétations de type Hélio-Clim (http://www.soda-is.com/eng/helioclim/index.html).
Ci-dessous, l'abstract proposé pour la conférence :

"Considering the grid manager's point of view, needs in terms of prediction of intermittent energy like the PV resource can be distinguished according to the considered horizon: the resource that will be available on the following days (d+1, d+2 et d+3), the next day by hourly step (h+24), during the next hour (h+1), and in the five next minutes (m+5). These horizons allow understanding the various aspects of the prediction: the medium term, the short term and the very short term. Through this work, we have identified some methodologies using time series models for the prediction horizon of global radiation and for the PV power. What we present here is a comparison of different predictors developed and tested to propose a hierarchy."

http://www.swc2013.org/site/ http://www.swc2013.org/site/
swc2013_template_full_paper_1_cv.pdf SWC2013_TEMPLATE FULL_PAPER-1-CV.pdf  (386.68 Ko)



Dimanche 24 Février 2013

Prévision de la ressource solaire en milieu insulaire

cyril Voyant


Participation à des journées thématiques organisées par le Laboratoire PIMENT de l'université de La Réunion

Le programme de ces journées abordera les thématiques liées aux méthodes de prévision de la ressource solaire en milieu insulaire.
Afin de concevoir des outils de prévision de la ressource solaire adaptés au cas insulaire (et en particulier à La Réunion), des séances de travail seront conduites avec les experts internationaux afin d’établir la meilleure stratégie permettant de participer à l’atteinte de l’objectif d’autonomie énergétique des iles (et en particulier à l’horizon 2025 pour La Réunion).
La mise en œuvre de ces outils de prévision passe par l’étude des modèles statistiques et numériques alliée à une instrumentation spécifique (imagerie du ciel « Total Sky Imager » ; exploitation des données satellitaires).
Dans le cadre de ce colloque, les invités utilisant déjà ces différents modes de prévision vont collaborer avec les membres de l’équipe de recherche PIMENT travaillant sur le sujet.
Des méthodologies seront alors discutées afin d’élaborer une stratégie globale pour la prévision de la ressource solaire en milieu insulaire.

Auront des représentants :

-State University of New-York
-Université d’OLDENBURG
-LMD Ecole Polytechnique
-EDF RD
-Université de Corse
-Université Antilles Guyane
-Agence Martiniquaise de l’Energie
-CEA-INES
-Météo FRANCE
-Université de La Réunion




Hybrid methodology for hourly global radiation forecasting in Mediterranean area

Cyril Voyant, Marc Muselli, Christophe Paoli, Marie Laure Nivet


Renewable Energy (IF3)
Renewable Energy (IF3)
Comme nous l'avions remarqué à plusieurs reprises, le mixage (ou l'hybridation) de méthodes prédictives est souvent intéressant. Nous présentons ici une méthodologie de prédiction basée sur les deux forecasters les plus utilisés en modélisation de séries temporelles : les méthodes linéaires ARMA et non-linéaires ANN. Un constat ressort de cette étude, concernant les journées à forte occurrence nuageuse, les approches de type non-linéaire sont les plus aptes à modéliser l'intermittence, alors que pour les journées plus ensoleillées, l'approche linéaire est souvent plus appropriée.

Pour plus de détails, voici l'abstract de cette publication : The renewable energies prediction and particularly global radiation forecasting is a challenge studied by a growing number of research teams. This paper proposes an original technique to model the insolation time series based on combining Artificial Neural Network (ANN) and Auto-Regressive and Moving Average (ARMA) model. While ANN by its non-linear nature is effective to predict cloudy days, ARMA techniques are more dedicated to sunny days without cloud occurrences. Thus, three hybrids models are suggested: the first proposes simply to use ARMA for 6 months in spring and summer and to use an optimized ANN for the other part of the year; the second model is equivalent to the first but with a seasonal learning; the last model depends on the error occurred the previous hour. These models were used to forecast the hourly global radiation for five places in Mediterranean area. The forecasting performance was compared among several models: the 3 above mentioned models, the best ANN and ARMA for each location. In the best configuration, the coupling of ANN and ARMA allows an improvement of more than 1%, with a maximum in autumn (3.4%) and a minimum in winter (0.9%) where ANN alone is the best.

https://www.researchgate.net/publication/233385988_Hybrid_methodology_for_hourly_global_radiation_forecasting_in_Mediterranean_area https://www.researchgate.net/publication/233385988_Hybrid_methodology_for_hourly_global_radiation_forecasting_in_Mediterranean_area



Jeudi 15 Novembre 2012

PV output power fluctuations smoothing: The MYRTE platform experience
C. Darras, M. Muselli, P. Poggi, C. Voyant, J.-C. Hoguet, F. Montignac


Papier accepté dans "International Journal of Hydrogen Energy"
Dans la continuité de son travail doctoral, C. Darras continue de perfectionner le logiciel ORIENTE sensé à terme piloter la station MYRTE (plan PV+stockage).
Le nouvel add-on présenté dans ce papier permet de comprendre et de valider le lissage des fluctuations de la ressource solaire par le biais des piles à combustible. Fort de cette première validation, nous sommes actuellement en cours de développement de la partie prédiction de la ressource. Les ANN et les méthodologies hybrides que l'on a développé dans le cadre des dernières publications (energy, solar energy, etc.) vont permettre de finaliser cette étude.



https://www.researchgate.net/publication/232252038_PV_output_power_fluctuations_smoothing_The_MYRTE_platform_experience/file/d912f507da0f554558.pdf https://www.researchgate.net/publication/232252038_PV_output_power_fluctuations_smoothing_The_MYRTE_platform_experience/file/d912f507da0f554558.pdf



Samedi 15 Septembre 2012 | Commentaires (1)
1 2 3 4