Article
Geosciences, Multidisciplinary
Yan Ji, Xiefei Zhi, Luying Ji, Yingxin Zhang, Cui Hao, Ting Peng
Summary: This study used deep-learning-based models to improve probabilistic precipitation forecasting over China and found that the deep NN model outperformed traditional methods and raw ensemble in predicting extreme precipitation events. The size of the training samples was found to have a significant impact on the results, and the forecast performance was better in central China.
FRONTIERS IN EARTH SCIENCE
(2022)
Article
Management
Nicole Ludwig, Siddharth Arora, James W. Taylor
Summary: Probabilistic forecasting of electricity demand is important for efficient management and operations of energy systems. The relationship between load and weather is complex and nonlinear, making load modeling using weather challenging. This study focuses on using weather ensemble predictions to model load in Great Britain. The ensembles are post-processed using ensemble model output statistics and ensemble copula coupling to improve their accuracy. The proposed approach outperforms other methods that do not use weather information or employ raw weather ensembles.
JOURNAL OF THE OPERATIONAL RESEARCH SOCIETY
(2023)
Article
Energy & Fuels
Tilmann Gneiting, Sebastian Lerch, Benedikt Schulz
Summary: Probabilistic solar forecasts can take the form of predictive probability distributions, ensembles, quantiles, or interval forecasts. The state-of-the-art approaches utilize input from numerical weather prediction models and apply statistical and machine learning methods for post-processing. This study proposes a probabilistic benchmark based on deterministic forecast and introduces new methods that merge statistical techniques with modern neural networks for post-processing.
Article
Green & Sustainable Science & Technology
Gokhan Mert Yagli, Dazhi Yang, Dipti Srinivasan
Summary: This article presents a method for generating and post-processing ensemble solar forecasts using satellite data. The method employs a dropout neural network model and various post-processing techniques, and it has been demonstrated to be effective in improving the quality of solar forecasts. The findings of this study are valuable to stakeholders in the power system industry.
RENEWABLE & SUSTAINABLE ENERGY REVIEWS
(2022)
Article
Energy & Fuels
Hadrien Verbois, Yves-Marie Saint-Drenan, Alexandre Thiery, Philippe Blanc
Summary: The share of solar power in the global and local energy mixes has significantly increased in the past decade, leading to a rise in interest for solar power forecasting. Numerical Weather Prediction (NWP) models and post-processing algorithms are the most popular methods for day-ahead forecasts. However, comparing results across different studies is challenging due to variations in datasets, metrics, and cross-validation methods. This study proposes a rigorous benchmark of solar NWP post-processing models using an open dataset spanning 6 years and 7 locations. The results demonstrate the systematic benefits of using large predictor sets with proper regularization, as well as the superior performance of more complex algorithms such as neural networks and gradient boosting in terms of mean square error. Support vector regression, a more parsimonious algorithm, performs better in terms of mean absolute error. The study highlights the importance of considering systematic ranking when evaluating forecasting models and emphasizes that no single model is superior in all situations.
Article
Energy & Fuels
Xianglong Li, Longfei Ma, Ping Chen, Hui Xu, Qijing Xing, Jiahui Yan, Siyue Lu, Haohao Fan, Lei Yang, Yongqiang Cheng
Summary: The paper presents a probabilistic prediction model of solar irradiance based on XGBoost, which utilizes historical data to train a point prediction model and generates probability prediction intervals under different confidence levels using kernel density estimation. Experimental results demonstrate that this method has better accuracy and is suitable for engineering practice.
Article
Green & Sustainable Science & Technology
Dazhi Yang, Gokhan Mert Yagli, Dipti Srinivasan
Summary: The paper introduces a state-of-the-art probabilistic solar forecasting method that effectively addresses the challenges of reflecting high-frequency fluctuations and changing uncertainty in solar energy systems, benefiting real-time stochastic simulations on a large scale.
RENEWABLE & SUSTAINABLE ENERGY REVIEWS
(2022)
Article
Energy & Fuels
Guanjun Liu, Hui Qin, Qin Shen, Hao Lyv, Yuhua Qu, Jialong Fu, Yongqi Liu, Jianzhong Zhou
Summary: A new spatiotemporal probabilistic prediction model combining a convolutional shared weight long short-term memory network and deep ensemble method is proposed in this paper to address solar radiation probabilistic prediction issues. The model optimizes uncertainty estimation and reliability of probabilistic prediction results by adjusting network structure and employing proper scoring rules. The model is shown to provide accurate point predictions, reasonable prediction intervals, and reliable probabilistic prediction results for a whole area in the evaluation against five state-of-the-art models and seven evaluation indicators.
Article
Meteorology & Atmospheric Sciences
Maria Lakatos, Sebastian Lerch, Stephan Hemri, Sandor Baran
Summary: Ensemble forecasts are an important step in weather forecasting as they account for uncertainties in future atmospheric conditions. However, they often lack accuracy and require post-processing. Multivariate post-processing aims to capture spatial and temporal correlations among different dimensions to improve forecast accuracy. In this study, different non-parametric multivariate approaches are compared, and results show that multivariate post-processing significantly enhances forecast skill compared to raw and independently calibrated forecasts.
QUARTERLY JOURNAL OF THE ROYAL METEOROLOGICAL SOCIETY
(2023)
Article
Chemistry, Multidisciplinary
Domingos S. de O. Jr Jr Santos, Paulo S. G. de Mattos Neto, Joao F. L. de Oliveira, Hugo Valadares Siqueira, Tathiana Mikamura Barchi, Aranildo R. Lima, Francisco Madeiro, Douglas A. P. Dantas, Attilio Converti, Alex C. Pereira, Jose Bione de Melo Filho, Manoel H. N. Marinho
Summary: Solar irradiance forecasting is crucial for renewable energy generation, as it enhances the planning and operation of photovoltaic systems. Traditional single models may underperform due to inappropriate selection, misspecification, or random fluctuations. This research proposes a heterogeneous ensemble dynamic selection model that outperforms single models in terms of accuracy.
APPLIED SCIENCES-BASEL
(2022)
Article
Meteorology & Atmospheric Sciences
Kirien Whan, Jakob Zscheischler, Alexander I. Jordan, Johanna F. Ziegel
Summary: Statistical post-processing plays a crucial role in providing accurate weather forecasts and early warnings. This study compared various methods and found that MBCn and OA are more effective in capturing dependence structures compared to ECC and the Schaake Shuffle.
WEATHER AND CLIMATE EXTREMES
(2021)
Article
Computer Science, Information Systems
Dayeong So, Jinyeong Oh, Subeen Leem, Hwimyeong Ha, Jihoon Moon
Summary: This study introduces HYTREM, a hybrid tree-based ensemble learning model designed to enhance the efficiency of solar power generation systems. The model showed superior performance compared to state-of-the-art models, demonstrating the lowest mean absolute error, root mean square error (RMSE), and normalized RMSE values across two regions.
Article
Meteorology & Atmospheric Sciences
Zohreh Javanshiri, Maede Fathi, Seyedeh Atefeh Mohammadi
Summary: This study explores the use of ensemble forecasting in probabilistic weather forecasting, emphasizing the importance of statistical post-processing to improve forecast quality. Results show that BMA and EMOS-CSG techniques are successful in enhancing WRF ensemble forecasts, with BMA being more accurate, skillful, and reliable, while EMOS-CSG method exhibits better resolution in predicting high-precipitation events.
METEOROLOGICAL APPLICATIONS
(2021)
Article
Engineering, Multidisciplinary
Zhongju Wang, Long Wang, Chao Huang, Xiong Luo
Summary: A hybrid multi-modal ensemble learning model is proposed for short-term solar irradiance forecasting, which combines historical observations and sky images. The model utilizes historical observations to extract temporal characteristics and uses ground-based sky images for cloud cover information. The Extreme Gradient Boosting (XGBoost) algorithm is employed to capture the relationship between input features and future observations. Experimental results show that the proposed hybrid model outperforms benchmarking methods in terms of forecasting accuracy.
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS
(2023)
Article
Multidisciplinary Sciences
Gobu Balraj, Aruldoss Albert Victoire, S. Jaikumar, Amalraj Victoire
Summary: A novel approach combining VMD and FTSVM models with deep learning mechanism is proposed to forecast the solar PV output power. The results show that the proposed model outperforms the existing models in terms of prediction accuracy and performance.
Article
Computer Science, Artificial Intelligence
Agnes Baran, Sebastian Lerch, Mehrez El Ayari, Sandor Baran
Summary: Accurate forecasting of total cloud cover is important for various sectors, and statistical calibration using machine learning methods can significantly improve forecast skill. Adding precipitation forecast data can further enhance predictive performance.
NEURAL COMPUTING & APPLICATIONS
(2021)
Article
Environmental Sciences
Sandor Baran, Patricia Szokol, Marianna Szabo
Summary: In recent years, ensemble weather forecasting has become routine at major weather prediction centers, but can often be underdispersive and biased, requiring postprocessing. This study proposes a novel EMOS model for calibrating wind speed ensemble forecasts, utilizing a truncated GEV distribution. Experimental results confirm the advantageous properties of the new model.
Article
Meteorology & Atmospheric Sciences
Benedikt Schulz, Sebastian Lerch
Summary: This study provides a comprehensive review and systematic comparison of eight statistical and machine learning methods for ensemble postprocessing of wind gust forecasts. The results show that incorporating additional meteorological predictor variables beyond wind gusts leads to significant improvements in forecast skill. A flexible framework of locally adaptive neural networks is proposed, which outperforms all benchmark postprocessing methods and learns physically consistent relations associated with the diurnal cycle and the evening transition of the planetary boundary layer.
MONTHLY WEATHER REVIEW
(2022)
Article
Meteorology & Atmospheric Sciences
William E. Chapman, Luca Delle Monache, Stefano Alessandrini, Aneesh C. Subramanian, F. Martin Ralph, Shang-Ping Xie, Sebastian Lerch, Negin Hayatbini
Summary: This study examines the use of deep-learning methods for postprocessing weather predictions, finding that these methods can generate reliable and accurate probabilistic forecasts. The deep-learning methods outperform the calibrated Global Ensemble Forecast System in specific meteorological events. Additionally, the results show that deep-learning methods learn quickly and require a relatively short hindcast dataset to compete with the Global Ensemble Forecast System.
MONTHLY WEATHER REVIEW
(2022)
Article
Energy & Fuels
Kaleb Phipps, Sebastian Lerch, Maria Andersson, Ralf Mikut, Veit Hagenmeyer, Nicole Ludwig
Summary: This paper examines multiple strategies for applying ensemble post-processing to probabilistic wind power forecasts and finds that post-processing the final wind power ensemble improves forecast performance, while only post-processing the weather ensembles does not necessarily lead to increased forecast performance.
Article
Meteorology & Atmospheric Sciences
Maria Lakatos, Sebastian Lerch, Stephan Hemri, Sandor Baran
Summary: Ensemble forecasts are an important step in weather forecasting as they account for uncertainties in future atmospheric conditions. However, they often lack accuracy and require post-processing. Multivariate post-processing aims to capture spatial and temporal correlations among different dimensions to improve forecast accuracy. In this study, different non-parametric multivariate approaches are compared, and results show that multivariate post-processing significantly enhances forecast skill compared to raw and independently calibrated forecasts.
QUARTERLY JOURNAL OF THE ROYAL METEOROLOGICAL SOCIETY
(2023)
Article
Energy & Fuels
Tilmann Gneiting, Sebastian Lerch, Benedikt Schulz
Summary: Probabilistic solar forecasts can take the form of predictive probability distributions, ensembles, quantiles, or interval forecasts. The state-of-the-art approaches utilize input from numerical weather prediction models and apply statistical and machine learning methods for post-processing. This study proposes a probabilistic benchmark based on deterministic forecast and introduces new methods that merge statistical techniques with modern neural networks for post-processing.
Article
Meteorology & Atmospheric Sciences
Marianna Szabo, Estibaliz Gascon, Sandor Baran
Summary: The study investigates the predictive performance of the censored shifted gamma (CSG) ensemble model output statistic (EMOS) approach for statistical postprocessing using dual-resolution ensemble forecasts over Europe. The results show that compared with the raw ensemble combinations, EMOS postprocessing significantly improves forecast skill and there is no statistical difference in skill between any of the analyzed mixtures of dual-resolution combinations.
WEATHER AND FORECASTING
(2023)
Article
Geosciences, Multidisciplinary
Riccardo Silini, Sebastian Lerch, Nikolaos Mastrantonas, Holger Kantz, Marcelo Barreiro, Cristina Masoller
Summary: This study uses multiple linear regression and a machine learning algorithm to improve the forecast of the European Centre for Medium-Range Weather Forecasts model for the Madden-Julian Oscillation (MJO). The results show that both methods improve the MJO prediction, with machine learning outperforming linear regression. The biggest improvement is seen in the prediction of the MJO geographical location and intensity.
EARTH SYSTEM DYNAMICS
(2022)
Article
Meteorology & Atmospheric Sciences
Sandor Baran, Agnes Baran
Summary: Wind power is the second largest energy source in the EU, but accurate wind power predictions are crucial for successful integration into the grid due to its volatility. Calibration of wind speed ensemble forecasts using a flexible machine learning approach results in the best overall performance.
Article
Energy & Fuels
Siddharth Sradhasagar, Omkar Subhasish Khuntia, Srikanta Biswal, Sougat Purohit, Amritendu Roy
Summary: In this study, machine learning models were developed to predict the bandgap and its character of double perovskite materials, with LGBMRegressor and XGBClassifier models identified as the best predictors. These models were further employed to predict the bandgap of novel bismuth-based transition metal oxide double perovskites, showing high accuracy, especially in the range of 1.2-1.8 eV.
Article
Energy & Fuels
Wei Shuai, Haoran Xu, Baoyang Luo, Yihui Huang, Dong Chen, Peiwang Zhu, Gang Xiao
Summary: In this study, a hybrid model based on numerical simulation and deep learning is proposed for the optimization and operation of solar receivers. By applying the model to different application scenarios and considering multiple performance objectives, small errors are achieved and optimal structure parameters and heliostat scales are identified. This approach is not only applicable to gas turbines but also heating systems.
Article
Energy & Fuels
Mubashar Ali, Zunaira Bibi, M. W. Younis, Muhammad Mubashir, Muqaddas Iqbal, Muhammad Usman Ali, Muhammad Asif Iqbal
Summary: This study investigates the structural, mechanical, and optoelectronic properties of the BaCuF3 fluoroperovskite using the first-principles modelling approach. The stability and characteristics of different cubic structures of BaCuF3 are evaluated, and the alpha-BaCuF3 and beta-BaCuF3 compounds are found to be mechanically stable with favorable optical properties for solar cells and high-frequency UV applications.
Article
Energy & Fuels
Dong Le Khac, Shahariar Chowdhury, Asmaa Soheil Najm, Montri Luengchavanon, Araa mebdir Holi, Mohammad Shah Jamal, Chin Hua Chia, Kuaanan Techato, Vidhya Selvanathan
Summary: A novel recycling system is proposed in this study to decompose and reclaim the constituent materials of organic-inorganic perovskite solar cells (PSCs). By utilizing a one-step solution process extraction approach, the chemical composition of each layer is successfully preserved, enabling their potential reuse. The proposed recycling technique helps mitigate pollution risks, minimize waste generation, and reduce recycling costs.
Article
Energy & Fuels
Peijie Lin, Feng Guo, Xiaoyang Lu, Qianying Zheng, Shuying Cheng, Yaohai Lin, Zhicong Chen, Lijun Wu, Zhuang Qian
Summary: This paper proposes an open-set fault diagnosis model for PV arrays based on 1D VoVNet-SVDD. The model accurately diagnoses various types of faults and is capable of identifying unknown fault types.