Bayesian Deep Learning-based Confidence-aware Solar Irradiance Forecasting System

Lee, HyunYong; Lee, Byung-Tak

doi:10.1109/ictc.2018.8539601

Cited by 11 publications

(17 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We compare with state-of-the-art label noise algorithms on multiple real-world open-source web-crawled datatsets including corrupted versions of the miniImageNet [39] and Stanford Cars [16] datasets provided by Jiang et al [12], the mini-WebVision dataset [23], and the Clothing1M [44] dataset. We observe that noisy OOD samples can be leveraged to improve network generalization by enforcing dynamically softening of labels tending to a uniform distribution [20] rather than discarding them. This paper's contributions are:…”

Section: Introductionmentioning

confidence: 99%

Addressing out-of-distribution label noise in webly-labelled data

Albert

Ortego²,

Arazo³

et al. 2022

2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

View full text Add to dashboard Cite

A recurring focus of the deep learning community is towards reducing the labeling effort. Data gathering and annotation using a search engine is a simple alternative to generating a fully human-annotated and human-gathered dataset. Although web crawling is very time efficient, some of the retrieved images are unavoidably noisy, i.e. incorrectly labeled. Designing robust algorithms for training on noisy data gathered from the web is an important research perspective that would render the building of datasets easier. In this paper we conduct a study to understand the type of label noise to expect when building a dataset using a search engine. We review the current limitations of stateof-the-art methods for dealing with noisy labels for image classification tasks in the case of web noise distribution. We propose a simple solution to bridge the gap with a fully clean dataset using Dynamic Softening of Out-of-distribution Samples (DSOS), which we design on corrupted versions of the CIFAR-100 dataset, and compare against state-of-the-art algorithms on the web noise perturbated MiniImageNet and Stanford datasets and on real label noise datasets: WebVision 1.0 and Clothing1M. Our work is fully reproducible https://git.io/JKGcj.

show abstract

Section: Introductionmentioning

confidence: 99%

Addressing out-of-distribution label noise in webly-labelled data

Albert

Ortego²,

Arazo³

et al. 2022

2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

View full text Add to dashboard Cite

show abstract

“…Deep Learning (DL) is an advanced Machine Learning approach that has been widely applied in many fields and has shown great performance for many problems such as image processing [5], [6], computer vision [7], [8], natural language processing [9], [10] and time series prediction [11], [12]. Also, the DL approaches have provided good accuracy for energy systems such as solar irradiance forecasting [13], [14] and wind speed prediction [15], [16]. Recently, DL approaches have been widely applied to predict the quantity of energy to be consumed.…”

Section: Introductionmentioning

confidence: 99%

Impact of Input Data Structure on Convolutional Neural Network Energy Prediction Model

Toumia¹,

Hassine²

2022

IJACSA

View full text Add to dashboard Cite

Energy demand continues to increase with no prospect of slowing down in the future. This increase is caused by several sociological and economical factors such as population growth, urbanization and technological developments. In view of this growth, it becomes crucial to predict energy consumption for a more accurate management and optimization. Nevertheless, consumption estimation is a complex task due to consumer behaviour fluctuation and weather alterations. Several efforts were proposed in the literature. Almost, all of them focused on improving the prediction model to increase the accuracy of the results. They use the LSTM (Long-Short Term Memory) model to reflect the temporal dependencies between historical data despite its spatial and temporal complexities. The main contribution in this paper is a novel and simple Convolutional Neural Network energy prediction model based on input data structure enhancement. The main idea is to adjust the structure of the input data instead of using a more complicated deep learning model for better performance. The proposed model was implemented, tested using real data and compared to existing ones. The obtained results showed that the proposed data structure has a great influence on the model performance measurement.

show abstract

“…Various structures of LSTM networks in [19], and [20] have been proposed for solar power forecasting. Lee et al [21], and Alzahrani et al [22] have demonstrated the superior performance of deep models over conventional techniques for solar irradiance estimation. The authors in [10] have shown that the ensemble approaches in PV output forecasting increase the precision and efficiency of models compared with individual models by integrating linear and non-linear techniques.…”

Section: Introductionmentioning

confidence: 99%

An Ensemble Approach for Multi-Step Ahead Energy Forecasting of Household Communities

et al. 2021

View full text Add to dashboard Cite

This paper addresses the estimation of household communities' overall energy usage and solar energy production, considering different prediction horizons. Forecasting the electricity demand and energy generation of communities can help enrich the information available to energy grid operators to better plan their short-term supply. Moreover, households will increasingly need to know more about their usage and generation patterns to make wiser decisions on their appliance usage and energy-trading programs. The main issues to address here are the volatility of load consumption induced by the consumption behaviour and variability in solar output influenced by solar cells specifications, several meteorological variables, and contextual factors such as time and calendar information. To address these issues, we propose a predicting approach that first considers the highly influential factors and, second, benefits from an ensemble learning method where one Gradient Boosted Regression Tree algorithm is combined with several Sequenceto-Sequence LSTM networks. We conducted experiments on a public dataset provided by the Ausgrid Australian electricity distributor collected over three years. The proposed model's prediction performance was compared to those by contributing learners and by conventional ensembles. The obtained results have demonstrated the potential of the proposed predictor to improve short-term multi-step forecasting by providing more stable forecasts and more accurate estimations under different day types and meteorological conditions.

show abstract

Bayesian Deep Learning-based Confidence-aware Solar Irradiance Forecasting System

Cited by 11 publications

References 14 publications

Addressing out-of-distribution label noise in webly-labelled data

Addressing out-of-distribution label noise in webly-labelled data

Impact of Input Data Structure on Convolutional Neural Network Energy Prediction Model

An Ensemble Approach for Multi-Step Ahead Energy Forecasting of Household Communities

Contact Info

Product

Resources

About