A Transformer-based multimodal-learning framework using sky images for ultra-short-term solar irradiance forecasting

Applied Energy, Год журнала: 2023, Номер 342, С. 121160 - 121160

Опубликована: Май 1, 2023

Язык: Английский

---

A Review of State-of-the-Art and Short-Term Forecasting Models for Solar PV Power Generation

Energies, Год журнала: 2023, Номер 16(14), С. 5436 - 5436

Опубликована: Июль 17, 2023

Accurately predicting the power produced during solar generation can greatly reduce impact of randomness and volatility on stability grid system, which is beneficial for its balanced operation optimized dispatch reduces operating costs. Solar PV depends weather conditions, such as temperature, relative humidity, rainfall (precipitation), global radiation, wind speed, etc., it prone to large fluctuations under different conditions. Its characterized by randomness, volatility, intermittency. Recently, demand further investigation into uncertainty short-term prediction effective use in many applications renewable energy sources has increased. In order improve predictive accuracy output develop a precise model, authors used algorithms system. Moreover, since forecasting an important aspect optimizing control systems electricity markets, this review focuses models generation, be verified daily planning smart addition, methods identified reviewed literature are classified according input data source, case studies examples proposed analyzed detail. The contributions, advantages, disadvantages probabilistic compared. Finally, future proposed.

Язык: Английский

---

Vision transformer models to measure solar irradiance using sky images in temperate climates

Applied Energy, Год журнала: 2024, Номер 362, С. 122967 - 122967

Опубликована: Март 13, 2024

Solar Irradiance measurements are critical for a broad range of energy systems, including evaluating performance ratios photovoltaic as well forecasting power generation. Using sky images to evaluate solar irradiance, allows low-cost, low-maintenance, and easy integration into Internet-of-things network, with minimal data loss. This work demonstrates that vision transformer-based machine learning model can produce accurate irradiance estimates based on sky-images without any auxiliary being used. The training utilizes 17 years global horizontal, diffuse direct data, high precision pyranometer pyrheliometer sun-tracked system; in-conjunction from standard lens fish-eye camera. learns attend relevant features the highly both horizontal (RMSE =52 W/m2) = 31 W/m2). compares model's wide field view all-sky camera shows transformer works best images. For normal convolutional architectures perform similarly convolution-based architecture showing an advantage RMSE 155 W/m2.

Язык: Английский

---

A multi-modal deep clustering method for day-ahead solar irradiance forecasting using ground-based cloud imagery and time series data

Energy, Год журнала: 2025, Номер unknown, С. 135285 - 135285

Опубликована: Март 1, 2025

Язык: Английский

---

A Review of State-of-the-art and Short-Term Forecasting Models for Solar PV Power Generation

Опубликована: Май 23, 2023

Accurately predicting the power of solar generation can greatly reduce impact randomness and volatility on stability grid system, which is beneficial for balanced operation optimized dispatch reduces operating costs. Solar PV depends weather conditions, are prone to large fluctuations under different conditions. Its characterized by randomness, intermittency. Recently, demand further investigation effective use uncertainty short-term prediction has been getting increasing attention in many application renewable energy sources. In order improve predictive accuracy output develop a precise model, authors worked algorithms system. Moreover, since forecasting one important aspects optimizing control systems electricity markets, this review focuses models generation, be verified daily planning smart addition, methods reviewed literature classified according input data source used accurate models, case studies examples proposed analyzed detail. The contributions, advantages disadvantages probabilistic compared. Finally, future proposed.

Язык: Английский

---

Short-term solar radiation forecast using total sky imager via transfer learning

Energy Reports, Год журнала: 2022, Номер 9, С. 819 - 828

Опубликована: Ноя. 18, 2022

Ground-based sky cameras, which capture hemispherical images, have been extensively used for localized monitoring of clouds. This paper proposes a short-term forecasting approach based on transfer learning using Total Sky-Imager (TSI) images the Southern Great Plains (SGP) site obtained from Atmospheric Radiation Measurement (ARM) dataset. An accurate estimation solar irradiance TSI is key energy generation and optimal consumption planning. We make use deep neural network architectures such as AlexNet ResNet-101 to extract underlying convolution features then train an ensemble model forecast radiation. demonstrate performance proposed by showcasing best worst cases. Thus, significantly reduces time resources required modeling outperform with reference another state-of-art technique at different lead times.

Язык: Английский

---

An ultra-short-term forecast method for photovoltaic power generation based on ResNet

Journal of Physics Conference Series, Год журнала: 2025, Номер 2935(1), С. 012002 - 012002

Опубликована: Янв. 1, 2025

Abstract Predicting photovoltaic power generation is important for enhancing the operation and management of systems, as well boosting their electricity production efficiency. An ultra-short-term prediction model based on ResNet proposed to address current issues unsatisfactory ground-based cloud image performance unfavorable marginalization deployment. After analyzing processing sky image, network used feature extraction. Historical data input fused with results Finally, predicted. The result experiment showed that can effectively extract information predicting generation.

Язык: Английский

---

Deep Learning-Based Image Regression for Short-Term Solar Irradiance Forecasting on the Edge

Electronics, Год журнала: 2022, Номер 11(22), С. 3794 - 3794

Опубликована: Ноя. 18, 2022

Photovoltaic (PV) power production is characterized by high variability due to short-term meteorological effects such as cloud movements. These have a significant impact on the incident solar irradiance in PV parks. In order control park performance, researchers focused Computer Vision and Deep Learning approaches perform forecasting using sky images. Motivated task of improving control, current work introduces Image Regression Module, which produces values from images image processing methods Convolutional Neural Networks (CNNs). With objective enhancing performance CNN models estimation forecasting, we propose an method based sun localization. Our findings show that proposed can consistently improve accuracy produced all our study, reducing Root Mean Square Error up 10.44 W/m2 for MobileNetV2 model. indicate future applications utilize CNNs should identify position produce more accurate values. Moreover, integration edge-oriented Field-Programmable Gate Array (FPGA) towards smart real-time emphasizes their advantages.

Язык: Английский

---

Spectral and RGB analysis of the light climate and its ecological impacts using an all-sky camera system in the Arctic

Applied Optics, Год журнала: 2023, Номер 62(19), С. 5139 - 5139

Опубликована: Март 22, 2023

The ArcLight observatory provides an hourly continuous time series of all-sky images providing light climate data (intensity, spectral composition, and photoperiod) from the Arctic (Svalbard at 79°N). Until recently, no complete annual relevant for biological processes has been provided high because insufficient sensitivity commercial sensors during Polar Night. set up is unique, as it both corresponding integrated irradiance in visible part solar electromagnetic spectrum (EPAR ). Here we present a further development diel-annual dynamics 2020 partitioned into red, green, blue parts illustrate their relation to weather conditions, sun moon trajectories. We show that there variation between RGB proportions throughout year, with showing greatest variation, which dependent on conditions (i.e., cloud cover). provide example impact these variations using vivo Chl a-specific absorption coefficients diatoms (mean six low acclimated northern-Arctic bloom-forming species) model total algal (AQ ) fraction quanta used by Photosystem II (AQPSII) (O2 production) bands potential impacts photoreceptor response, suggesting periods where repair maintenance functions dominate activity absence appreciable levels red or green light. method here can be applied response worldwide give localized ecological models AQ.

Язык: Английский

---

Multidimensional Feature Extraction Based Minutely Solar Irradiance Forecasting Method Using All-Sky Images

IEEE Transactions on Industry Applications, Год журнала: 2024, Номер 60(3), С. 4494 - 4504

Опубликована: Март 5, 2024

With the increasing penetration of photovoltaic (PV) power generation in grid, minutely irradiance prediction, which is basis PV has become very important. Aiming at problems extracting feature redundancy, weakening key area features, and high cloud-sky misidentification rate current this paper proposes a solar forecasting method based on multidimensional extraction using all-sky image, beneficial to achieve higher accuracy forecasting. First, improved clustering-boundary correction algorithm used identify cloud sky pixels classify images into four types. Then capture sub domain that will cover sun future dynamically according results displacement vector calculation extract local features overall as convolutional neural network (CNN) image RGB matrix, respectively. Finally combined with meteorological factors historical respectively construct mapping models for types prediction ten-minute scale. Compared benchmarks, mean absolute percentage error (MAPE) proposed reduced by 0.21%, 20.21%, 2.53%, 5.30% types: clear sky, block clouds, thin thick The can be widely plants equipped imagers provide data support optimal operation maintenance plant.

Язык: Английский

---