National Renewable Energy Laboratory Research Articles

NUMERICAL INVESTIGATION OF THERMAL STRATIFICATION EFFECTS ON WIND TURBINE WAKE FLOW

Numerical investigation of thermal stratification effects on wind turbine wake flow with and without yawed conditions was conducted in this paper. The three-dimensional wind turbine model from National Renewable Energy Laboratory (NREL) Phase VI experiment was adopted in all the simulations. Model validation was done by using three commonly used turbulence models: realizable <i>k-ε</i> model, shear-stress transport (SST) <i>k-ω</i> model, and large eddy simulation (LES) model. Simulation results were compared with experimental data in the literature. The realizable <i>k-ε</i> model was selected for further simulation as it showed a good balance in achieving accurate results and lowering computational cost. Followed by the validation, wind turbine wake flow with and without yawed effects under different thermal stratification conditions were simulated. Wake flow parameters were discussed and analyzed, followed by suggestions of wind turbine layouts based on different operation conditions to efficiently harness wind energy.

Risk Minimization in Scale-Up of Biomass and Waste Carbon Upgrading Processes.

Improving the odds and pace of successful biomass and waste carbon utilization technology scale-up is crucial to decarbonizing key industries such as aviation and materials within timelines required to meet global climate goals. In this perspective, we review deficiencies commonly encountered during scale-up to show that many nascent technology developers place too much focus on simply demonstrating that technologies work in progressively larger units ("profit") without expending enough up-front research effort to identify and derisk roadblocks to commercialization (collecting "information") to inform the design of these units. We combine this conclusion with economic and timeline data collected from technology scale-up and piloting operations at the National Renewable Energy Laboratory (NREL) to motivate a more scientific, risk-minimized approach to biomass and waste carbon upgrading scale-up. Our proposed approach emphasizes maximizing information collection in the smallest, most agile, and least expensive experimental setups possible, emulating the mentality embraced by R&D across the petrochemical industry. Key points are supported by examples of successful and unsuccessful scale-up efforts undertaken at NREL and elsewhere. We close by showing that the U.S. national laboratory system is uniquely well equipped to serve as a hub to facilitate effective scale-up of promising biomass and waste carbon upgrading technologies.

Uncertainty quantification for dynamic responses of offshore wind turbine based on manifold learning

Offshore wind turbines (WTs) are crucial in offshore wind energy development. However, the dynamic responses of WTs are subject to significant uncertainties which are usually not properly considered. To the end, this paper proposes an efficient method for quantifying the uncertainties in WTs' dynamic responses based on cumulative distribution function (CDF)-manifold learning. First, a probabilistic model is developed to represent the environmental parameters and sampling for aerodynamic-hydraulic-servo-elastic simulations. Then, the CDF is obtained by statistically analyzing the simulated data. To tackle the higher dimensionality resulting from discretizing the CDF, a manifold learning-based approach is subsequently proposed to reduce its dimensionality and obtain a manifold space. Furthermore, a mapping relation is established between the environmental parameters and the low-dimensional data to efficiently obtain the response CDF under different environmental parameters, leading to the construction of a probability box (P-box) model. To demonstrate the effectiveness of the proposed method, the National Renewable Energy Laboratory (NREL) 5 MW offshore WT on an Offshore Code Comparison Collaboration (OC3) monopile is selected as a case study and analyzed accordingly. The results show P-box models of seven WT responses and validate the effectiveness of the proposed method.

(Invited) Mulitjunction III-V Semiconductors for Photo-Electrochemical Hydrogen Production: Recent Progress in Efficiency, Durability, and Cost

Photo-electrochemical (PEC) hydrogen production is a route to carbon-free fuel that combines light absorption and water electrolysis into one device. Challenging targets for solar-to-hydrogen (STH) efficiency, durability, and semiconductor absorber costs must be realized for this approach to be economically competitive with other hydrogen production pathways. Research on PEC hydrogen production at the National Renewable Energy Laboratory (NREL) has focused on III-V semiconductor absorber materials because their tunable bandgaps, high photon conversion efficiencies, and ability to make multi-junction structures have resulted in the highest STH efficiencies yet demonstrated. This talk will focus on recent NREL results showing progress toward PEC hydrogen production efficiency, durability, and cost targets. Specifically, we have been able to improve theoretically achievable STH efficiency through bandgap engineering in multijunction devices. Photon management strategies to close the gap between theoretical and measured STH efficiency will also be discussed. Durability remains a significant challenge for III-V materials in a PEC environment. We will describe modifications to the semiconductor surface and the electrolyte constitution that can make the photoabsorber less susceptible to photocorrosion during operation. Finally, we will present preliminary results on III-V PEC materials synthesized via hydride vapor phase epitaxy, an alternative route to high-efficiency photoelectrodes that could potentially have much lower costs than the metalorganic vapor phase epitaxy method traditionally used. These improvements have yielded III-V PEC devices that are close to the STH efficiency target (17% vs. 25%) while the demonstrated durability and cost are still considerably far from their target values.

(Invited) Benchmarking and Predicting the Risks of Commercial Li-Ion Cells Undergoing Thermal Runaway

The performance of commercial Li-ion batteries increases year after year. New mechanical designs of cells as well as changing electrolyte and active material compositions continually push performance boundaries forward. With such changes, the risks associated with thermal runaway of cells change too and a challenge persists in recording and benchmarking risks to determine whether increased performance comes at the expense of increased risk in the event of thermal runaway. The Battery Failure Databank provides a record of thermal and mass ejection behaviors of past and present commercial cells and provides a valuable resource for benchmarking behaviors and determining the predictability of risks associated with thermal runaway.A Fractional Thermal Runaway Calorimeter (FTRC), developed by the NASA Johnson Space Center in collaboration with the National Renewable Energy Laboratory (NREL), distinctly measures heat emitted from the body of commercial cells and heat that is ejected from the cell. The FTRC also facilitates in-situ high-speed radiography to link phenomena occurring inside the cell with external thermal measurements. The FTRC was used for all tests recorded in the Battery Failure Databank and high-speed radiographs accompany each thermal dataset. Having thermal, mass ejection, and internal dynamics behaviors from hundreds commercial cells under mechanical and thermal abuse conditions facilitates statistical analyses of cell behaviors and evaluation of the efficacy of machine learning methods for predicting risks associated with thermal runaway.Using data from the Battery Failure Databank, the following three topics will be covered in this talk: Correlations between heat generation and mass ejection: The total heat output from cells showed a strong correlation with the quantity of material ejected. An explanation for this based on distinct heat measurements from ejected and non-ejected materials will be discussed. Potential causes of outlier cells that produce exceptionally high heat: Upon conducting numerous repeat tests for each cell, outlier cells that produce exceptionally high quantities of heat were observed. High-speed radiography data revealed that internal dynamic phenomena can lead to extreme variations in thermal and ejection behavior observed externally. The predictability of thermal runaway outcomes using machine learning methods and limited data: With limited data on thermal behavior, mass ejection, and cell specifications, the accuracy of zero- and one-shot approaches to predicting the variation and distribution of heat output during thermal runaway of a new cell type, based on training data in the Battery Failure Databank, will be discussed. The analyses and open-access resources provided with this work are expected to provide a foundation for benchmarking the behavior and predictability of risks associated with thermal runaway of past and present commercial Li-ion cells. In addition to providing a valuable resource for designing safe battery systems, it is expected that this data will be imperative as we evaluate next generation cell designs and assess their respective failure mechanisms and associated risks.

A new similarity criterion for wind tunnel testing of floating wind turbines achieving dynamic aerodynamic characteristics similarity

Floating offshore wind turbines have rapidly advanced in recent years, prompting numerous experiments. Small-scale turbines require blade redesign with low Reynolds number airfoils due to the scale effect. Consequently, the chord length of these blades no longer corresponds to the length scale, leading to dissimilar dynamic aerodynamic phenomena during platform movement. To address this, a new criterion is proposed for wind tunnel tests of floating offshore wind turbines. The criterion incorporates reduced frequency correction and optimal tip speed ratio (TSR) matching. It ensures hydrodynamic similarity through Froude number similarity and avoids excessive low wind speed through wind speed correction. Optimal TSR matching ensures the model and prototype turbines operate within similar power curve intervals. Reduced frequency corrects the platform sine motion parameters, providing the model turbine with a dynamic aerodynamic effect similar to the prototype. The National Renewable Energy Laboratory’s 5 MW wind turbine is chosen as the prototype, and the model wind turbine blades are designed using NACA 4412 airfoil and the criterion of thrust coefficient similarity. The sub-module of the FAST, AeroDyn is employed to validate the new criterion for different surge motions between the NREL 5 MW and the model. Under the new similarity criterion, the model turbine exhibits a relative root mean square error of 4.61 % and 1.54 % in power coefficient and thrust coefficient, a significant improvement compared to the current similarity criterion's relative root mean square error of 24.74 % and 8.36 %. The results reveal a circular trend variation in both thrust coefficient and power coefficient with position. Notably, the new criterion demonstrates a closer alignment with the dynamic aerodynamic characteristics of the prototype when contrasted with the current similarity criterion. Furthermore, this model design, tailored for the NREL 5 MW turbine, can be applied to model experiments with various turbine configurations. It has been successfully verified for the IEA 15 MW turbine, exhibiting promising simulation results. The proposed similarity criterion is suitable for wind tunnel testing, addressing the problem that the current similarity criterion cannot reflect the dynamic aerodynamic characteristics of prototype wind turbines. Additionally, it enables experiments on different prototype configurations using a single model turbine, reducing wind tunnel testing costs.

A Point-Interval Forecasting Method for Wind Speed Using Improved Wild Horse Optimization Algorithm and Ensemble Learning

With the implementation of the green development strategy and the “double carbon goal”, as an important energy for sustainable development, wind power has been widely researched and vigorously developed across the world. Wind speed prediction has a major impact on the grid dispatching of wind power connection. Most current studies only focus on the deterministic prediction of wind speed. However, the traditional deterministic forecast only provides the single wind speed prediction results and cannot meet the diverse demands of dispatchers. To bridge the gap, a wind speed point-interval forecasting method is proposed that utilizes empirical wavelet transform, an improved wild horse optimization algorithm, a multi-predictor, and improved kernel density estimation. This method decomposes the wind speed sequence into stationary subsequences through empirical wavelet transform, and then optimizes three basic learners with completely different learning mechanisms to form an ensemble model using the modified wild horse optimization algorithm. Finally, the uncertainty is analysed using an improved kernel density estimation. The datasets of three sites from America’s national renewable energy laboratory are used for comparison experiments with other models, and the predictions are discussed from different angles. The simulation results demonstrate that the model can produce high-precision deterministic results and high-quality probabilistic results. The reference information the model provides can be extremely valuable for scheduling operators.

Design Methodology of Wind Turbine Rotor Models Based on Aerodynamic Thrust and Torque Equivalence

Limited by scaling effects, the physical model tests of FOWTs (floating offshore wind turbines) cannot simulate the aerodynamic loads on rotors correctly. To solve this problem, the real-time hybrid model tests in wind tunnels were developed and provided a feasible solution for the aerodynamic simulation. To perform the wind tunnel tests, the design of aerodynamic equivalent rotor models is most critical. In this study, an innovative methodology of aerodynamic equivalent design for the wind turbine rotors is developed based on GA (genetic algorithm). The NREL (National Renewable Energy Laboratory) 5 MW and DTU (Technical University of Denmark) 10 MW rotors are employed for the case studies to validate the proposed methodology. According to the results, the model-scale aerodynamic thrust performance can be accurately matched with the prototype in the entire region between cut-in and cut-out wind speeds, which allows the rotor model to provide correct thrust at different wind speeds. The variance of the aerodynamic torque with the wind speeds for the developed model is also in good agreement with the prototype, which could be beneficial for the design of the model-scale active pitch control strategy. Moreover, the applicability of the fitness functions of GA is discussed.

Comparative study on the dynamic responses of monopile and jacket-supported offshore wind turbines considering the pile-soil interaction in transitional waters

In this study, simplified finite element models for monopile and jacket offshore wind turbines, referred to as MOWT and JOWT, respectively, were established for dynamic response analysis using the X-SEA program, based on the National Renewable Energy Laboratory's 5 MW OWT model. First, using five design constraints, an optimization design was conducted for the substructures and foundations of both MOWT and JOWT located in transitional waters to ensure a fair comparison of the dynamic performances of the two structures. Next, putting special attention to the natural frequency, tower top displacement and member stress, two kinds of OWT models with fixed and flexible bases subjected to combined wind, wave and current loads were analyzed using the pile-soil interaction module of X-SEA to further explore the influence of the foundation flexibility on dynamic responses. Subsequently, a parametric study was conducted to evaluate the sensitivity of the dynamic responses to the changes in environmental conditions, soil properties and pile embedded lengths. Finally, a comparison of the dynamic performances, parameter sensitivities, and total masses of the MOWT and JOWT was performed to determine the optimal fixed OWT support structure for the 50 m water depth of the southwest sea area of South Korea.

CFD model of refuelling through the entire equipment of a hydrogen refuelling station

This paper aims at the development and validation of a computational fluid dynamic (CFD) model for simulations of the refuelling process through the entire equipment of the hydrogen refuelling station (HRS). The absence of such models hinders the design of inherently safer refuelling protocols for an arbitrary combination of HRS equipment, hydrogen storage parameters, and environmental conditions. The CFD model is validated against the complete process of refuelling lasting 195s in Test No.1 performed by the National Renewable Energy Laboratory (NREL). The test equipment includes high-pressure tanks of HRS, pressure control valve (PCV), valves, pipes, breakaway, hose, and nozzle all the way up to three onboard tanks. The model accurately reproduced hydrogen temperature and pressure through the entire line of HRS equipment. A standout feature of the CFD model, distinguishing it from simplified models, is the capability to predict temperature non-uniformity in onboard tanks, a crucial factor with significant safety implications.

Reliability of simplified wind fields and structure models for the response simulation of parked offshore wind turbines

Over the past few decades, offshore wind turbines (OWTs) have been popular in the wind engineering and renewable energy fields. Numerous studies have investigated the dynamic responses of OWTs by numerical simulation. However, the different numerical models gave different results, making it difficult to compare and select models for different design and research purposes. In this study, six models including the main wind field and structure model were built based on the National Renewable Energy Laboratory 5-MW OWT, and the dynamic responses were simulated to compare and quantify the effects of the models on the response of the parked OWT. The results show that a uniform distribution of the wind field is too conservative for estimating the dynamic responses of the parked OWT. The structural response calculated using a uniformly distributed wind field is twice as large as those with the spatial distribution and coherence of the wind speed. The wind profile significantly affected the response of the parked OWT, particularly on the standard deviation of the displacement. When estimating the dynamic response of the tower of a parked OWT, the simplified model, in which the blades and wind load are simplified as a concentrated mass and load, is relatively reliable. The maximum displacement at the tower top of the simplified model is close to that of the detailed model. However, a detailed model is necessary to obtain the blade responses and accurate power spectral density of the structural response. In addition, the wind-field model significantly affected the spectral characteristics of the dynamic response of the blades.

Bioclimatic design recommendations for novel cluster analysis-based mapping for humid climates with altitudinal gradient variations

Proper planning of urban environments in construction projects requires an accurate understanding of the prevailing climates in an area. However, the need for sufficient climate records in the Global South countries can make this characterization difficult. This study proposes using satellite information and cluster analysis to define the climatic behavior in Colombia, specifically in areas with intertropical climates with altitudinal gradient impact. The methodology used the statistical analysis of hourly temperature and relative humidity data for recent typical meteorological year (TMY) files. These data were obtained from the National Renewable Energy Laboratory (NREL). Subsequently, cluster analysis was applied to group regions with similar climatic characteristics. Standardization of the climatic variables ensured that all contributed equally to the research. The study results present an atlas with nine different climatic strategy zones, turning Colombia's climate into eight main clusters, each with specific characteristics and recommendations for the passive design of buildings. Identifying these zones made it possible to define the distribution of bioclimatic strategies in the different thermal floors of the region studied. This approach provides a framework for urban planners and city officials to develop climate-responsive building design guidelines adapted to the specific climatic conditions of each zone.

A deep learning-based transformer model for photovoltaic fault forecasting and classification

According to the US-based National Renewable Energy Lab (NREL), solar energy losses due to faults were 3.5 % in 2004, which increased to 17.5 % in 2018. Therefore, the fault prediction mechanism will enable PV practitioners to reduce losses effectively, enhancing the solar system's efficiency and power output. This paper proposes a deep learning-based Transformer model for robust fault prediction in photovoltaic. Transformer uses attention mechanism that considers data points as a language units “word” and learn dependencies among them to predict upcoming data points. Unlike other forecasting algorithms, our proposed approach does not rely on previous trends. In case of PV faults, trends do not exist. The proposed algorithm utilizes rate of change of solar cell parameters for establishing a trend to forecast faults, enabling proactive fault mitigation. It also classifies faults with different severity levels to identify the level of predictive maintenance required. The proposed approach is extensively evaluated using MATLAB on datasets of several faults with low, medium, and high severity levels. The proposed Transformer model achieves a forecasting mean average error (MAE) of 0.09377. Performance of the proposed forecasting and classification algorithm is compared with existing machine learning-based regression and classification techniques such as KNN, SVM, and NN, where proposed approach outperforms state-of-the-art approaches.

Highly energy efficient housing can reduce peak load and increase safety under beneficial electrification

Climate change is driving urgent investments in decarbonization. One core decarbonization strategy is to electrify energy services that currently directly use fossil fuels, because electricity can be generated from zero greenhouse gas energy resources. Shifting fossil-based services to electricity, however, requires a major expansion of electricity supply and increases dependence on electricity for critical services. Home heating is a particular challenge, especially in very cold climates. Unserved heating loads can be fatal. Electrified heating is expected to drive peak loads (and thus overall grid size) due to high coincident and nondeferrable loads. This study shows that highly efficient housing presents an opportunity to simultaneously protect people and structurally reduce peak load, reducing the need for electricity supply infrastructure while increasing people’s resilience to weather extremes. This study uses seven building efficiency scenarios from the National Renewable Energy Laboratory’s End Use Saving Shapes to investigate the impact of residential building efficiency on grid size in 2050, using the example of Pierre, South Dakota as a very cold weather location that might also experience substantial new housing demand due to climate-induced human mobility. We find that the deepest efficiency electrification scenario we investigate reduces peak demand by about half relative to low-efficiency electrification. Costs of about $3900/kilowatt (kW) peak load reduction are competitive with the cost of new decarbonized supplies capable of meeting peak load, though building efficiency costs are usually privatized while supply expansion costs are distributed across ratepayers. Decarbonization scenarios suggest the US grid might need to expand by a factor of 5–8 in the next 25 years: extremely rapid growth will be needed regardless, but targets might not be reachable with inefficient end users. Residential building efficiency presents an urgent opportunity to reduce peak demand and provide safer and more resilient housing.

Implications of uncertainty in technology cost projections for least-cost decarbonized electricity systems

Plans for decarbonized electricity systems rely on projections of highly uncertain future technology costs. We use a stylized model to investigate the influence of future cost uncertainty, as represented by different projections in the National Renewable Energy Laboratory 2021 Annual Technology Baseline dataset, on technology mixes comprising least-cost decarbonized electricity systems. Our analysis shows that given the level of future cost uncertainty as represented by these projections, it is not possible to predict with confidence which technologies will play a dominant role in future least-cost carbon emission-free energy systems. Successful efforts to reduce costs of individual technologies may or may not lead to system cost reductions and widespread deployments, depending on the success of cost-reduction efforts for competing and complementary technologies. These results suggest a portfolio approach to reducing technology costs. Reliance on uncertain cost breakthroughs risks costly outcomes. Iterative decision-making with learning can help mitigate these risks.

Regridding uncertainty for statistical downscaling of solar radiation

Abstract. Initial steps in statistical downscaling involve being able to compare observed data from regional climate models (RCMs). This prediction requires (1) regridding RCM outputs from their native grids and at differing spatial resolutions to a common grid in order to be comparable to observed data and (2) bias correcting RCM data, for example via quantile mapping, for future modeling and analysis. The uncertainty associated with (1) is not always considered for downstream operations in (2). This work examines this uncertainty, which is not often made available to the user of a regridded data product. This analysis is applied to RCM solar radiation data from the NA-CORDEX (North American Coordinated Regional Climate Downscaling Experiment) data archive and observed data from the National Solar Radiation Database housed at the National Renewable Energy Lab. A case study of the mentioned methods over California is presented.

Integration of a physics-based direct normal irradiance (DNI) model to enhance the National Solar Radiation Database (NSRDB)

The National Solar Radiation Database (NSRDB) is an extensively used dataset that furnishes satellite-retrieved solar resource data across the United States and an expanding list of other countries. Although the NSRDB uses a physical model to compute global horizontal irradiance (GHI), it currently employs an empirical approach based on surface observations to estimate cloudy-sky direct normal irradiance (DNI). Recently, a new physics-based approach, known as the Fast All-sky Radiation Model for Solar applications with DNI (FARMS-DNI), was developed to improve DNI forecasting. FARMS-DNI integrates direct and scattered solar radiances within the circumsolar region, resulting in improved day-ahead forecasting of DNI by incorporating it into the Weather Research and Forecasting model with Solar extensions (WRF-Solar). This study incorporates FARMS-DNI into the NSRDB algorithm to produce high-spatiotemporal-resolution DNI data from satellite data. The accuracy of the NSRDB based on FARMS-DNI is analyzed using surface observations from 19 sites situated within the National Oceanic and Atmospheric Administration (NOAA) Surface Radiation Budget (SURFRAD) and Solar Radiation (SOLRAD) networks, the University of Oregon (UO) network, the U.S. Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) network, and at the National Renewable Energy Laboratory (NREL). The results demonstrate that FARMS-DNI reduces the significant overestimation of DNI in the conventional NSRDB at all surface sites, particularly in cloud overcast conditions classified using both satellite retrievals and surface observations. Consequently, this new model can effectively improve the overall accuracy of the NSRDB. The results also suggest that further improvement of DNI estimates at individual time steps, however, requires advanced satellite techniques and precise identification of clouds and retrieval of cloud properties.

Short-term wind power forecasting based on multivariate/multi-step LSTM with temporal feature attention mechanism

Precision enhancement for short-term wind power forecasting can alleviate negative impact of the forecasting results on wind power generation. Due to complexities and nonlinearities among factors and facets in wind power, it is essential to achieve reliable and stable power generation via the long short-term memory (LSTM) forecasting. To this purpose, multi-task temporal feature attention (MTTFA) based LSTM, namely MTTFA-LSTM, is proposed for multivariate/multi-step wind power forecasting with historical power and meteorological data, in which task-sharing and task-specifying layers are designed for task co-features extracting and task specifics discriminating, respectively. More specifically, in the task-sharing layer, multi-dimensional inputs are fed into LSTM to extract long-term trends, while in the task-specifying layer, one-dimensional convolution operations extract temporal features hidden in each and all time steps. Furthermore, an attention mechanism is adopted to adaptively tune weights for temporal features. Finally, the proposed model is leveraged to cope with different short-term wind power forecasting (SWPF) problems based on the national renewable energy laboratory’s (NREL) wind power data. Simulation results show that the proposed MTTFA-LSTM achieves persistent excellent forecasting accuracy, comparing its backbone STL model, TFA-LSTM as well as the benchmark MTL models in the same setting, which indicate that the complex and non-linear interdependencies among multi-dimensional data can be well depicted by the proposed model.

Variable cross-sectional effect on bi-directional blades–tower–soil–structure dynamic interaction on offshore wind turbine subject to wind–wave loads

BackgroundThis study introduces a numerical model designed to simulate interactions occurring between a wind turbine's tower and the surrounding soil, as well as between the nacelle, blades, and the surrounding environment. This simulation accounts for both fore–aft and side-to-side movements. To describe these interactions, the model leverages the Euler–Lagrange equations. It calculates wave loads utilizing the Morison equation, with wave data generated based on the JONSWAP spectrum. Furthermore, aerodynamic loads are determined using the blade element moment theory, and the wind spectrum is generated using the Von Karman turbulence model. The tower is represented as a variable cross-sectional beam, employing a two-noded Euler beam element with two degrees of freedom: transverse displacement and rotation, and utilizing Hermite polynomial shape functions.ResultsIn a comparative analysis against experimental data, this modified model demonstrates significant enhancements in accurately reproducing the dynamic behavior of wind turbines with variable cross-sectional towers, outperforming models that approximate the tower with a constant cross section. Our findings reveal that the modified model achieves a remarkable improvement of 15% in replicating the tower's dynamic response when compared to the constant cross-sectional models. As a case study, a 5 MW monopile wind turbine with a flexible foundation, specifically the one provided by the National Renewable Energy Laboratory (NREL), is employed to simulate its dynamic response.ConclusionsThis research presents a robust numerical model for simulating wind turbine behavior in various environmental conditions. The incorporation of variable cross-sectional tower representation significantly improves the model's accuracy, making it a valuable tool for assessing wind turbine dynamics. The study's findings highlight the importance of considering tower flexibility in wind turbine simulations to enhance their real-world applicability.

Photovoltaic power prediction based on sky images and tokens-to-token vision transformer

Photovoltaic (PV) power generation has high uncertainties due to the randomness and imbalance nature of solar energy and meteorological parameters. Hence, accurate PV power forecasts are essential in the operation of PV power plants (PVPP) for short-term dispatches and power generation schedules. In this paper, a new deep neural network structure based on vision transformer is proposed to combine sky images and Tokens-To-Token（T2T） for photovoltaic power prediction. The method uses an incremental tokenization module to aggregate neighboring image patches into tokens, which capture the local structural information of the clouds. Then, an efficient T2T-ViT backbone network is used to extract the global attentional relationships of the tokens for power prediction. In order to evaluate the performance of the proposed model, the method was compared with several deep learning architectures such as ResNet and GoogleNet on a dataset collected by the National Renewable Energy Laboratory in Colorado, USA. The results of power prediction were analysed using training loss, prediction error, and linear regression, and they show that the proposed method achieves higher prediction accuracy and lower error compared to the existing methods, especially in short- and ultra-short-term prediction. The paper demonstrates the potential of applying Transformer models to computer vision tasks for renewable energy forecasting. The results show that the proposed method achieves higher prediction accuracy and lower error than several deep learning architectures, such as ResNet and GoogleNet, especially in short- and ultra-short-term prediction.