Accéder directement au contenu Accéder directement à la navigation
Communication dans un congrès

Photovoltaic power prediction using a recurrent neural network RNN

Abstract : The intermittent nature of solar energy creates a significant challenge for the optimization and planning of future smart grids. In order to reduce intermittency, it is very important to accurately predict Photovoltaic (PV) power generation. This work proposes a new prediction method based on the Recurrent Neural Network (RNN) for accurately predicting the yield of photovoltaic power generation systems. Our study used a Longe Short-Term Memory (LSTM) architecture. The LSTM approach can store information over time, which is valuable for time series prediction. The proposed prediction method is evaluated using real PV energy in Lille, France. Firstly, all solar time series data are divided into three main parts: 70% of the data are used to train the neural network, 20% of the data are used for verification and the other data are used for testing. The proposed prediction method has a good prediction quality in very short term (one-hour), which proves the reliability and cost-effectiveness of this method.
Type de document :
Communication dans un congrès
Liste complète des métadonnées
Contributeur : Louise DESSAIVRE Connectez-vous pour contacter le contributeur
Soumis le : lundi 11 avril 2022 - 10:31:26
Dernière modification le : mercredi 21 septembre 2022 - 15:24:39


  • HAL Id : hal-03636732, version 1


Mohamed Hamza Kermia, Dhaker Abbes, Jerome Bosche. Photovoltaic power prediction using a recurrent neural network RNN. 2020 6TH IEEE INTERNATIONAL ENERGY CONFERENCE (ENERGYCON), Sep 2020, Gammarth, Tunisia. pp.545-549. ⟨hal-03636732⟩



Consultations de la notice