A wavelet-LSTM model for short-term wind power forecasting using wind farm SCADA data

Expert Systems with Applications, Год журнала: 2024, Номер 247, С. 123237 - 123237

Опубликована: Янв. 19, 2024

Язык: Английский

---

Unveiling Latent Chemical Mechanisms: Hybrid Modeling for Estimating Spatiotemporally Varying Parameters in Moving Boundary Problems

Industrial & Engineering Chemistry Research, Год журнала: 2024, Номер 63(3), С. 1501 - 1514

Опубликована: Янв. 9, 2024

Hybrid modeling has gained substantial recognition due to its capacity seamlessly integrate machine learning methodologies while preserving the fundamental physical principles inherent in a model. Although these hybrid models have predominantly been applied temporal processes governed by ordinary differential equations, intricacies of various real-world systems, characterized diverse array processes, extend far beyond this domain. This study extensively investigates application techniques within context complex biological system regulated reaction–diffusion equation, specific type partial equation. The primary objective is tackle challenges associated with latent chemical mechanisms that are concealed from direct observation. proposed approach introduces framework synergizes neural networks mathematical methods estimate parameters exhibiting spatiotemporal variations category problems involving dynamic boundaries. Model training executed through back-propagation algorithm, adept at efficiently updating ensuring numerical stability. Subsequently, model employed models, yielding results validate proficiency precisely estimating variable diffusivity across both space and time as well fluctuations cell proliferation rates carrying density. mean squared error final output predictions 9.4 × 10–6 compared simple first-principles which 3.12 10–4.

Язык: Английский

---

Modeling based on machine learning to investigate flue gas desulfurization performance by calcium silicate absorbent in a sand bed reactor

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

Опубликована: Янв. 10, 2024

Abstract Flue gas desulfurization (FGD) is a critical process for reducing sulfur dioxide (SO 2 ) emissions from industrial sources, particularly power plants. This research uses calcium silicate absorbent in combination with machine learning (ML) to predict SO concentration within an FGD process. The collected dataset encompasses four input parameters, specifically relative humidity, weight, temperature, and time, incorporates one output parameter, which pertains the of . Six ML models were developed estimate parameters. Statistical metrics such as coefficient determination (R mean squared error (MSE) employed identify most suitable model assess its fitting effectiveness. random forest (RF) emerged top-performing model, boasting R 0.9902 MSE 0.0008. model's predictions aligned closely experimental results, confirming high accuracy. hyperparameter values RF found be 74 n_estimators, 41 max_depth, false bootstrap, sqrt max_features, 1 min_samples_leaf, absolute_error criterion, 3 min_samples_split. Three-dimensional surface plots generated explore impact variables on concentration. Global sensitivity analysis (GSA) revealed weight time significantly influence integration into modeling offers novel approach optimizing efficiency effectiveness this environmentally crucial

Язык: Английский

---

Low-Carbon Jujube Moisture Content Detection Based On Spectral Selection and Reconstruction

IEEE Internet of Things Journal, Год журнала: 2024, Номер 11(24), С. 38953 - 38964

Опубликована: Фев. 21, 2024

In recent years, the combination of hyperspectral imagery and deep learning has been widely used in agricultural Artificial Intelligence Things (AIoT), such as product quality assessment crop disease detection. However, this often comes at cost substantial computational power energy consumption. paper, we focused on data-efficient green computing for low-carbon jujube moisture content First, order to compress images capacity, a spectral selection algorithm based swarm intelligence was proposed screen necessary sensitive dimensions. Then, reconstruction model established realize selected bands from RGB image, aiming reduce high imaging. Finally, fusion data reconstructed image constructed efficient The experimental results show that 10 feature screened by method can adequately characterize water information jujube, outperforms other works with MRAE 0.1635. carbon emissions our are significantly lower than methods. Further, spectral-image achieves satisfactory detection result content, RMSE 0.0082. summary, selection, reconstruction, methods achieve precision while reducing emissions, which have important guidance sustainable AIoT applications.

Язык: Английский

---

Data-Driven Optimization of High-Dimensional Variables in Proton Exchange Membrane Water Electrolysis Membrane Electrode Assembly Assisted by Machine Learning

Industrial & Engineering Chemistry Research, Год журнала: 2024, Номер 63(3), С. 1409 - 1421

Опубликована: Янв. 11, 2024

The optimization of the membrane electrode assembly (MEA) is crucial for enhancing performance proton exchange water electrolysis. Nevertheless, achieving global all manufacturing parameters MEA poses challenges due to their high-dimensional complexity and limited experimental data. In this study, machine learning (ML) techniques were introduced tackle intricate engineering challenge. 58 MEAs fabricated tested construct a comprehensive database enriched with features ample This was achieved through data expansion method that involves altering operating temperature electrolyzer. XGBoost employed perform regression predictions on variables, remarkable coefficient determination (R2) value 0.99926. SHAP (SHapley Additive exPlanations) genetic algorithm applied model interpretation optimization, respectively. By utilizing insights provided by method, we could narrow decision variable dimensionality down 5 key results are comparable full-variable while notably reducing time costs 67.9%. Guided ML, globally optimized variables voltage only 1.828 V at 3 A cm–2. study presents an approach integrates intelligent data-driven methods optimization. contribution provides valuable into energy conversion storage technologies in chemical industry.

Язык: Английский

---

Digital twin-aided transfer learning for energy efficiency optimization of thermal spray dryers: Leveraging shared drying characteristics across chemicals with limited data

Applied Thermal Engineering, Год журнала: 2024, Номер 242, С. 122431 - 122431

Опубликована: Янв. 21, 2024

Язык: Английский

---

Achieving robustness in hybrid models: A physics-informed regularization approach for spatiotemporal parameter estimation in PDEs

Process Safety and Environmental Protection, Год журнала: 2024, Номер 204, С. 292 - 302

Опубликована: Фев. 2, 2024

Язык: Английский

---

An online transfer learning model for wind turbine power prediction based on spatial feature construction and system-wide update

Applied Energy, Год журнала: 2023, Номер 340, С. 121049 - 121049

Опубликована: Апрель 10, 2023

Язык: Английский

---

Leveraging Deep Learning to Strengthen the Cyber-Resilience of Renewable Energy Supply Chains: A Survey

IEEE Communications Surveys & Tutorials, Год журнала: 2024, Номер 26(3), С. 2146 - 2175

Опубликована: Янв. 1, 2024

Deep learning shows immense potential for strengthening the cyber-resilience of renewable energy supply chains. However, research gaps in comprehensive benchmarks, real-world model evaluations, and data generation tailored to domain persist. This study explores applying state-of-the-art deep techniques secure chains, drawing insights from over 300 publications. We aim provide an updated, rigorous analysis applications this field guide future research. systematically review literature spanning 2020-2023, retrieving relevant articles major databases. examine learning's role intrusion/anomaly detection, chain cyberattack detection frameworks, security standards, historical attack analysis, management strategies, architectures, cyber datasets. Our demonstrates enables anomaly by processing massively distributed data. highlight crucial design factors, including accuracy, adaptation capability, communication security, resilience adversarial threats. Comparing 18 attacks informs risk analysis. also showcase evaluating their relative strengths limitations applications. Moreover, our emphasizes best practices curation, considering quality, labeling, access efficiency, governance. Effective integration necessitates tuning guidance, generation. multi-dimensional motivates focused efforts on enhancing explanations, securing communications, continually retraining models, establishing standardized assessment protocols. Overall, we a roadmap progress leveraging potential.

Язык: Английский

---

Dynamic Neural Network Structure: A Review for Its Theories and Applications

IEEE Transactions on Neural Networks and Learning Systems, Год журнала: 2024, Номер 36(3), С. 4246 - 4266

Опубликована: Апрель 9, 2024

The dynamic neural network (DNN), in contrast to the static counterpart, offers numerous advantages, such as improved accuracy, efficiency, and interpretability. These benefits stem from network's flexible structures parameters, making it highly attractive applicable across various domains. As broad learning system (BLS) continues evolve, DNNs have expanded beyond deep (DL), orienting a more comprehensive range of Therefore, this review article focuses on two prominent areas where DNN rapidly developed: 1) DL 2) learning. This provides an in-depth exploration techniques related construction inference. Furthermore, discusses applications diverse domains while also addressing open issues highlighting promising research directions. By offering understanding DNNs, serves valuable resource for researchers, guiding them toward future investigations.

Язык: Английский

---