Novel model for medium to long term photovoltaic power prediction using interactive feature trend transformer

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Фев. 24, 2025

The stochastic and variable nature of power generated by photovoltaic (PV) systems can impact grid stability. Accurately predicting the output a solar PV generation system is crucial for addressing this challenge. While short-term prediction highly accurate, accuracy medium- to long-term predictions will face great challenges. In order improve medium prediction, unique hybrid deep learning model named interactive feature trend transformer (IFTformer) has been designed. Initially, isolated forest (DIF) local anomaly factor (LOF) are used construct parallel framework that serves as data preprocessing module, removing outliers from raw data. time series subsequently decomposed into seasonal components, which modelled separately independent study. Ultimately, predicted components with ProSparse Self-attention mechanism based on information interaction fitted multilayer perceptron (MLP) prediction. comprehensive experimental results show predictive performance IFTformer superior baseline models, normalised root mean square error (NRMSE) 3.64% absolute (NMAE) 2.44%. proposed in paper an effective approach mitigate outliers, enhance extraction ability, accuracy, generalizability robustness predictions, providing novel perspective methods methods.

Язык: Английский

---

Artificial intelligence-based prediction and analysis of the oversupply of wind and solar energy in power systems

Energy Conversion and Management, Год журнала: 2021, Номер 250, С. 114892 - 114892

Опубликована: Окт. 27, 2021

Язык: Английский

---

A Wind Energy Supplier Bidding Strategy Using Combined EGA-Inspired HPSOIFA Optimizer and Deep Learning Predictor

Energies, Год журнала: 2021, Номер 14(11), С. 3059 - 3059

Опубликована: Май 25, 2021

As the integration of large-scale wind energy is increasing into electricity grids, role suppliers should be investigated as a price-maker their participation would influence locational marginal price (LMP) electricity. The existing bidding strategies for supplier faces limitations with respect to potential cooperation, other competitors’ behavior, network loss, and uncertainty production (WP) balancing market (BMP). Hence, solve these problems, novel strategy (BS) power has been proposed in this paper. new algorithm, called evolutionary game approach (EGA) inspired hybrid particle swarm optimization improved firefly algorithm (HPSOIFA), handle issue. behavior suppliers, including conventional encoded one species obtain equilibrium where EGA can explore dynamically reasonable changes opponents. Each change exploited by HPSOIFA improve solutions. Moreover, deep learning namely belief network, implemented improving accuracy forecasting results considering WP BMP, revealed BMP modeled quantile regression (QR). Finally, Shapley value (SV) calculated estimate benefits cooperative suppliers. presented case studies have verified that established exhibit higher effectiveness.

Язык: Английский

---

Predicting solar radiation for space heating with thermal storage system based on temporal convolutional network-attention model

Applied Thermal Engineering, Год журнала: 2022, Номер 219, С. 119574 - 119574

Опубликована: Окт. 30, 2022

Язык: Английский

---

Fault diagnosis for PV system using a deep learning optimized via PSO heuristic combination technique

Electrical Engineering, Год журнала: 2023, Номер 105(4), С. 2287 - 2301

Опубликована: Март 30, 2023

Abstract A heuristic particle swarm optimization combined with Back Propagation Neural Network (BPNN-PSO) technique is proposed in this paper to improve the convergence and accuracy of prediction for fault diagnosis Photovoltaic (PV) array system. This works by applying ability deep learning classification find best solution search space. Some parameters are extracted from output PV be used identification purpose The results using back propagation neural network method only combination compared. algorithm converges after 350 steps while BP-PSO 250 training phase. BP algorithms about 87.8% achieved 95% right predictions. It was clearly shown that had better simulation as well

Язык: Английский

---

Improving the prediction of biochar production from various biomass sources through the implementation of eXplainable machine learning approaches

International Journal of Green Energy, Год журнала: 2024, Номер 21(12), С. 2771 - 2798

Опубликована: Март 14, 2024

Examining the game-changing possibilities of explainable machine learning techniques, this study explores fast-growing area biochar production prediction. The paper demonstrates how recent advances in sensitivity analysis methodology, optimization training hyperparameters, and state-of-the-art ensemble techniques have greatly simplified enhanced forecasting output composition from various biomass sources. argues that white-box models, which are more open comprehensible, crucial for prediction light increasing suspicion black-box models. Accurate forecasts guaranteed by these AI systems, also give detailed explanations mechanisms generating outcomes. For models to gain confidence processes enable informed decision-making, there must be an emphasis on interpretability openness. comprehensively synthesizes most critical features a rigorous assessment current literature relies authors' own experience. Explainable encourage ecologically responsible decision-making improving forecast accuracy transparency. Biochar is positioned as participant solving global concerns connected soil health climate change, ultimately contributes wider aims environmental sustainability renewable energy consumption.

Язык: Английский

---

A Future Direction of Machine Learning for Building Energy Management: Interpretable Models

Energies, Год журнала: 2024, Номер 17(3), С. 700 - 700

Опубликована: Фев. 1, 2024

Machine learning (ML) algorithms are now part of everyday life, as many technological devices use these algorithms. The spectrum uses is wide, but it evident that ML represents a revolution may change almost every human activity. However, for all innovations, comes with challenges. One the most critical challenges providing users an understanding how models’ output related to input data. This called “interpretability”, and focused on explaining what feature influences model’s output. Some have simple easy-to-understand relationship between output, while other models “black boxes” return without giving user information influenced it. lack this knowledge creates truthfulness issue when inspected by human, especially operator not data scientist. Building Construction sector starting face innovation, its scientific community working define best practices models. work intended developing deep analysis determine interpretable could be among promising future technologies energy management in built environments.

Язык: Английский

---

Optimizing electricity peak shaving through stochastic programming and probabilistic time series forecasting

Journal of Building Engineering, Год журнала: 2024, Номер 88, С. 109163 - 109163

Опубликована: Март 28, 2024

Язык: Английский

---

Solar Irradiance Forecasting with Natural Language Processing of Cloud Observations and Interpretation of Results with Modified Shapley Additive Explanations

Algorithms, Год журнала: 2024, Номер 17(4), С. 150 - 150

Опубликована: Апрель 2, 2024

Forecasting the generation of solar power plants (SPPs) requires taking into account meteorological parameters that influence difference between irradiance at top atmosphere calculated with high accuracy and tilted plane panel on Earth’s surface. One key factors is cloudiness, which can be presented not only as a percentage sky area covered by clouds but also many additional parameters, such type clouds, distribution across atmospheric layers, their height. The use machine learning algorithms to forecast retrospective data over long period formalising features; however, detailed information about cloudiness are normally recorded in natural language format. This paper proposes an algorithm for processing records convert them binary feature vector. Experiments conducted from real plant showed this increases short-term forecasts 5–15%, depending quality metric used. At same time, adding features makes model less transparent user, significant drawback point view explainable artificial intelligence. Therefore, uses additive explanation based Shapley vector interpret model’s output. It shown approach allows explain why it generates particular forecast, will provide greater level trust intelligent systems industry.

Язык: Английский

---

Machine learning analysis of thermophysical and thermohydraulic properties in ethylene glycol- and glycerol-based SiO2 nanofluids

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Июнь 27, 2024

The study investigates the heat transfer and friction factor properties of ethylene glycol glycerol-based silicon dioxide nanofluids flowing in a circular tube under continuous flux circumstances. This tackles important requirement for effective thermal management areas such as electronics cooling, automobile industry, renewable energy systems. Previous research has encountered difficulties enhancing performance while handling increased associated with nanofluids. conducted experiments Reynolds number range 1300 to 21,000 particle volume concentrations up 1.0%. Nanofluids exhibited superior coefficients values than base liquid values. highest enhancement was 5.4% 8.3% glycerol -based Nanofluid relative penalty ∼30% 75%, respectively. To model predict complicated, nonlinear experimental data, five machine learning approaches were used: linear regression, random forest, extreme gradient boosting, adaptive decision tree. Among them, tree-based performed well few errors, forest boosting models also highly accurate. findings indicate that these advanced can accurately anticipate nanofluids, providing dependable tool improving their use variety study's help design more cooling solutions improve sustainability

Язык: Английский

---