Comparison of machine learning methods for photovoltaic power forecasting based on numerical weather prediction

Renewable and Sustainable Energy Reviews, Journal Year: 2022, Volume and Issue: 161, P. 112364 - 112364

Published: March 23, 2022

The increase of the worldwide installed photovoltaic (PV) capacity and intermittent nature solar resource highlights importance power forecasting for grid integration technology. This study compares 24 machine learning models deterministic day-ahead based on numerical weather predictions (NWP), tested two-year-long 15-min resolution datasets 16 PV plants in Hungary. effects predictor selection benefits hyperparameter tuning are also evaluated. results show that two most accurate kernel ridge regression multilayer perceptron with an up to 44.6% forecast skill score over persistence. Supplementing basic NWP data Sun position angles statistically processed irradiance values as inputs a 13.1% decrease root mean square error (RMSE), which underlines selection. is essential exploit full potential models, especially less robust prone under or overfitting without proper tuning. overall best forecasts have 13.9% lower RMSE compared baseline scenario using linear regression. Moreover, only daily average 1.5% higher than scenario, demonstrates effectiveness even limited availability. this paper can support both researchers practitioners constructing data-driven techniques NWP-based forecasting.

Language: Английский

---

A review of solar forecasting, its dependence on atmospheric sciences and implications for grid integration: Towards carbon neutrality

Renewable and Sustainable Energy Reviews, Journal Year: 2022, Volume and Issue: 161, P. 112348 - 112348

Published: March 25, 2022

Language: Английский

---

Time series forecasting for hourly photovoltaic power using conditional generative adversarial network and Bi-LSTM

Energy, Journal Year: 2022, Volume and Issue: 246, P. 123403 - 123403

Published: Feb. 8, 2022

Language: Английский

---

Forecasting and uncertainty analysis of day-ahead photovoltaic power using a novel forecasting method

Applied Energy, Journal Year: 2021, Volume and Issue: 299, P. 117291 - 117291

Published: June 24, 2021

Language: Английский

---

Benefits of physical and machine learning hybridization for photovoltaic power forecasting

Renewable and Sustainable Energy Reviews, Journal Year: 2022, Volume and Issue: 168, P. 112772 - 112772

Published: July 14, 2022

Irradiance-to-power conversion is an essential step of state-of-the-art photovoltaic (PV) power forecasting, regardless the source and post-processing irradiance forecasts. The two distinct approaches for mapping forecasts to PV are physical data-driven, which can also be hybridized. contribution this paper twofold; first, it proposes a concept identifies best implementation hybrid machine learning irradiance-to-power method. Second, head-to-head comparison physical, methods performed operational day-ahead forecasting 14 plants in Hungary based on numerical weather prediction (NWP). To respect rule consistency but still obtain as complete picture possible, directives set, namely minimizing mean absolute error (MAE) root square (RMSE), separate sets optimized both directives. results reveal that years training data, method involves most physically-calculated predictors reduce MAE by 5.2% 10.4% compared, respectively, model chains without any considerations. important modeling steps separation transposition modeling, rest simulation left models significant increase errors. optimization found even case modeling; therefore, should become standard procedure practical applications. Finally, only beneficial at least one year while initial period operation plant, advised stay with modeling. guidelines recommendations help researchers practitioners design optimize their accuracy

Language: Английский

---

Trends and gaps in photovoltaic power forecasting with machine learning

Energy Reports, Journal Year: 2022, Volume and Issue: 9, P. 447 - 471

Published: Dec. 10, 2022

The share of solar energy in the electricity mix increases year after year. Knowing production photovoltaic (PV) power at each instant time is crucial for its integration into grid. However, due to meteorological phenomena, PV output can be uncertain and continuously varying, which complicates yield prediction. In recent years, machine learning (ML) techniques have entered world forecasting help increase accuracy predictions. Researchers seen great potential this approach, creating a vast literature on topic. This paper intends identify most popular approaches gaps discipline. To do so, representative part consisting 100 publications classified based different aspects such as ML family, location systems, number systems considered, features, etc. Via classification, main trends highlighted while offering advice researchers interested

Language: Английский

---

A Concise Overview on Solar Resource Assessment and Forecasting

Advances in Atmospheric Sciences, Journal Year: 2022, Volume and Issue: 39(8), P. 1239 - 1251

Published: Jan. 25, 2022

Abstract China’s recently announced directive on tackling climate change, namely, to reach carbon peak by 2030 and achieve neutrality 2060, has led an unprecedented nationwide response among the academia industry. Under such a directive, rapid increase in grid penetration rate of solar near future can be fully anticipated. Although radiation is atmospheric process, its utilization, as produce electricity, hitherto been handled engineers. In that, it thought important bridge two fields, sciences engineering, for common good neutrality. this überreview, all major aspects pertaining resource assessment forecasting are discussed brief. Given size topic at hand, instead presenting technical details, which would overly lengthy repetitive, overarching goal review comprehensively compile catalog some recent, not so papers, that interested readers explore details their own.

Language: Английский

---

Multi-timescale photovoltaic power forecasting using an improved Stacking ensemble algorithm based LSTM-Informer model

Energy, Journal Year: 2023, Volume and Issue: 283, P. 128669 - 128669

Published: Aug. 8, 2023

Language: Английский

---

Improved multistep ahead photovoltaic power prediction model based on LSTM and self-attention with weather forecast data

Applied Energy, Journal Year: 2024, Volume and Issue: 359, P. 122709 - 122709

Published: Feb. 2, 2024

Language: Английский

---

Short-term photovoltaic power forecasting using meta-learning and numerical weather prediction independent Long Short-Term Memory models

Renewable Energy, Journal Year: 2023, Volume and Issue: 216, P. 118997 - 118997

Published: July 13, 2023

Short-term photovoltaic (PV) power forecasting is essential for integrating renewable energy sources into the grid as it provides accurate and timely information on expected output of PV systems. Deep learning (DL) networks have shown promising results in this area, but depending weather conditions particularities each system, different DL architectures may perform best. This paper proposes a meta-learning method to improve one-hour-ahead deterministic forecasts systems by dynamically blending base multiple models learn under what model performs Four long short-term memory are used produce production without using numerical predictions, with objective enhance generalizability proposed solution. The accuracy meta-learner evaluated three rooftop Lisbon, Portugal. Results indicate that best at plants, can up 5% over most per plant 4.5% equal-weighted combination forecasts. These improvements statistically significant even larger during peak hours.

Language: Английский

---