A physics-guided machine learning framework for real-time dynamic wake prediction of wind turbines

Physics of Fluids, Journal Year: 2024, Volume and Issue: 36(3)

Published: March 1, 2024

Efficient and accurate prediction of the wind turbine dynamic wake is crucial for active control load assessment in farms. This paper proposes a real-time model turbines based on physics-guided neural network. The can predict instantaneous field under various operating conditions using only inflow speed as input. utilizes Taylor's frozen-flow hypothesis steady-state to convert parameters into network input features. A deep convolutional then maps these features desired snapshots, enabling predictions turbines. To train model, we generated approximately 255 000 flow snapshots single-turbine wakes large eddy simulation, covering different thrust coefficients yaw angles. was trained supervised learning method verified test set. results indicate that effectively characteristics, including meandering deflection yawed also assess both velocity center turbine. At coefficient 0.75, root mean square error predicted around 6.53%, while Pearson correlation reach 0.624. Furthermore, once trained, its accuracy does not decrease with increase time span.

Language: Английский

---

Digital twin enabled transition towards the smart electric vehicle charging infrastructure: A review

Sustainable Cities and Society, Journal Year: 2024, Volume and Issue: 108, P. 105479 - 105479

Published: May 3, 2024

Language: Английский

---

Performance evaluation and optimization of the cascade refrigeration system based on the digital twin model

Applied Thermal Engineering, Journal Year: 2024, Volume and Issue: 248, P. 123160 - 123160

Published: April 10, 2024

Language: Английский

---

Data-model-interactive enhancement-based Francis turbine unit health condition assessment using graph driven health benchmark model

Expert Systems with Applications, Journal Year: 2024, Volume and Issue: 249, P. 123724 - 123724

Published: March 20, 2024

Language: Английский

---

Innovative Horizons for Sustainable Smart Energy: Exploring the Synergy of 5G and Digital Twin Technologies

Process Integration and Optimization for Sustainability, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 14, 2025

Language: Английский

---

Wake field prediction of a wind farm based on a physics-informed neural network with different spatiotemporal prediction performance improvement strategies

Theoretical and Applied Mechanics Letters, Journal Year: 2025, Volume and Issue: unknown, P. 100577 - 100577

Published: Feb. 1, 2025

Language: Английский

---

Overview of Data-Driven Models for Wind Turbine Wake Flows

Journal of Marine Science and Application, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 8, 2025

Language: Английский

---

Enhanced Solar Photovoltaic System Management and Integration: The Digital Twin Concept

Solar, Journal Year: 2025, Volume and Issue: 5(1), P. 7 - 7

Published: March 6, 2025

The rapid acceptance of solar photovoltaic (PV) energy across various countries has created a pressing need for more coordinated approaches to the sustainable monitoring and maintenance these widely distributed installations. To address this challenge, several digitization architectures have been proposed, with one most recently applied being digital twin (DT) system architecture. DTs proven effective in predictive maintenance, prototyping, efficient manufacturing, reliable monitoring. However, while DT concept is well established fields like wind conversion monitoring, its scope implementation PV remains quite limited. Additionally, recent increased adoption autonomous platforms, particularly robotics, expanded management revealed gaps real-time needs. platforms can be redesigned ease such applications enable integration into broader network. This work provides system-level overview current trends, challenges, future opportunities within renewable systems, focusing on systems. It also highlights how advances artificial intelligence (AI), internet-of-Things (IoT), systems leveraged create digitally connected infrastructure that supports supply maintenance.

Language: Английский

---

Innovative sparse data reconstruction approaches for yawed wind turbine wake flow via data-driven and physics-informed machine learning

Physics of Fluids, Journal Year: 2025, Volume and Issue: 37(3)

Published: March 1, 2025

This paper explores innovative approaches for reconstructing the wake flow field of yawed wind turbines from sparse data using data-driven and physics-informed machine learning techniques. The estimation (WFE) integrates neural networks with fundamental fluid dynamics equations, providing robust interpretable predictions. method ensures adherence to essential principles, making it suitable reliable in energy applications. In contrast, (DDML-WFE) leverages techniques such as proper orthogonal decomposition extract significant features, offering computational efficiency reduced reconstruction costs. Both methods demonstrate satisfactory performance instantaneous under conditions. DDML-WFE maintains comparable even measurement resolution increased noise, highlighting its potential real-time turbine control. study employs a limited number points balance collection challenges while capturing characteristics. Future research will focus on optimizing control strategies farms by incorporating multi-scale modules advanced temporal prediction fields.

Language: Английский

---

A Data-Driven Algorithm for Dynamic Parameter Estimation of an Alkaline Electrolysis System Combining Online Reinforcement Learning and k-Means Clustering Analysis

Processes, Journal Year: 2025, Volume and Issue: 13(4), P. 1009 - 1009

Published: March 28, 2025

Determining the electrochemical, thermal, and mass transfer dynamics embedded in an alkaline electrolysis (AEL) system provides important information about application of ancillary services provided by hydrogen energy for elimination carbon emissions. Therefore, there is urgent need to develop methodologies evaluating key parameters, such as overvoltage coefficients, stack capacity, diaphragm thickness, permeability, accurately capture system’s fluctuating characteristics. However, limited lack superior sensor technology, some significant variables cannot be measured directly. In this context, comprehensively accurate parameters estimation strategy offer a novel alternative characterize corresponding intrinsic nature. This paper was motivated arduous challenge aims address large branching factors with irregular properties. Specifically, associated mathematical models reflecting transient operating terms heat transfer, are first established. Subsequently, k-means clustering analysis conducted deduce similarity distribution variables, which can function proxies separator distinguish working status. Furthermore, online reinforcement learning (RL), renowned its ability operate without extensive predefined datasets, employed conduct dynamic parameter estimation, thereby approximating robust nonlinear stochastic behaviors within AEL components. Finally, experimental results verify that proposed model achieves improvements errors compared existing methods (such EKF UKF). The enhancements 76.7%, 54.96%, 51.84%, 31% RMSE, NRMSE, PCC, MPE, respectively.

Language: Английский

---