A systematic review of machine learning methods applied to fuel cells in performance evaluation,durability prediction,and application monitoring

A fuel cell is a power generation device that directly converts chemical energy into electrical energy through chemical reactions; fuel cells are widely used in aerospace, electric vehicle, and small-scale stationary engine applications. The complex phenomena including mass/heat transfer, electrochemical reactions, and ion/electron conduction, can significantly affect the energy efficiency and durability of fuel cells, but are difficult to determine completely. Machine learning (ML) performs well in solving complex problems in engineering applications and scientific research. In this paper, a systematic review is conducted to explore ML methods, including traditional ML and deep learning (DL) methods, applied to fuel cells for performance evaluation (material selection, chemical reaction modeling, and polarization curves), durability prediction (state of health, fault diagnostics, and remaining useful life), and application monitoring. Then comparisons of traditional ML and DL methods are discussed, while the similarities and differences between ML and integrated physics simulations are also concluded. Eventually, the scope of ML methods applied to fuel cells is presented, and outlooks of future researches on ML applications in fuel cells are identified.     

Development and application of virtual nuclear power plant in digital society environment

Research and development of safe and efficient nuclear energy systems is imperative, since nuclear safety is the key issue in the development of nuclear energy, and it is also the premise of nuclear energy development learned from the painful lessons in history. Advanced numerical simulation can restore the complex physical processes as much as possible and predict system behavior and safety performance, and it facilitates accurate design and assessment. International research in nuclear safety simulation has developed from single physical phenomena simulation to coupling simulation with digital reactor and to comprehensive simulation with virtual nuclear power plant by integrating environmental and social information. Compared with digital reactor, virtual nuclear power plant pays more attention to the evolution of reactor accidents, such as large‐scale physical and social behavior simulation, which concerned the relationship between nuclear safety and environment, as well as the relationship between the nuclear safety and the public. FDS Team proposed “nuclear informatics” firstly by combining nuclear science and informatics and has developed the virtual nuclear power plant in digital society environment Virtual4DS, where many key technologies are developed under the guidance of nuclear informatics, such as integration cloud architecture, automatic accurate modeling, multiphysics coupling simulation, multidimensional information visualization and virtual simulation, and nuclear big data. Virtual4DS has been widely used in the nuclear power plant, nuclear weapons, well logging, etc.     

Statistical learning for wind power: A modeling and stability study towards forecasting

We focus on wind power modeling using machine learning techniques. We show on real data provided by the wind energy company Maïa Eolis that parametric models even following closely the physical equation relating wind production to wind speed are outperformed by intelligent learning algorithms. In particular, the CART‐Bagging algorithm gives very stable and promising results. Besides, as a step towards forecast, we quantify the impact of using deteriorated wind measures on the performances. We show also on this application that the default methodology to select a subset of predictors provided in the standard random forest package can be refined, especially when there exists among the predictors one variable, which has a major impact.     

Paradigm shift in urban energy systems through distributed generation: Methods and models

The path towards energy sustainability is commonly referred to the incremental adoption of available technologies, practices and policies that may help to decrease the environmental impact of energy sector, while providing an adequate standard of energy services. The evaluation of trade-offs among technologies, practices and policies for the mitigation of environmental problems related to energy resources depletion requires a deep knowledge of the local and global effects of the proposed solutions. While attempting to calculate such effects for a large complex system like a city, an advanced multidisciplinary approach is needed to overcome difficulties in modeling correctly real phenomena while maintaining computational transparency, reliability, interoperability and efficiency across different levels of analysis. Further, a methodology that rationally integrates different computational models and techniques is necessary to enable collaborative research in the field of optimization of energy efficiency strategies and integration of renewable energy systems in urban areas. For these reasons, a selection of currently available models for distributed generation planning and design is presented and analyzed in the perspective of gathering their capabilities in an optimization framework to support a paradigm shift in urban energy systems. This framework embodies the main concepts of a local energy management system and adopts a multicriteria perspective to determine optimal solutions for providing energy services through distributed generation.     

Horizontal Axis Wind Turbine Blade Aerodynamics in Experiments and Modeling

Turbine aerodynamics remains a challenging and crucial research area for wind energy. Blade aerodynamic forces responsible for power production must be augmented to maximize energy capture. At the same time, adverse aerodynamic loads that fatigue turbine components need to be mitigated to extend machine service life. Successful resolution of these conflicting demands and continued cost of energy reduction require accurate blade aerodynamic models. This, in turn, depends on clear physical understanding and reliable numerical modeling of rotational augmentation and dynamic stall, the two phenomena principally responsible for amplified turbine blade aerodynamic loads. The current work examines full-scale turbine blade aerodynamic measurements and current modeling methodologies to better understand the physical and numerical attributes that determine model performance     

A robust surrogate model of a solid oxide cell based on an adaptive polynomial approximation method

Surrogate models that predict the behaviors of solid oxide cells (SOCs) accurately at low computational cost are crucial to the control and optimization of SOC plants. Lumped physical models of SOCs, while widely used in such applications, lack accuracy because of neglected physical details. Data-driven models are the other options of surrogate models, which are proved to be more accurate because these models are identified directly from experiments or numerical simulations. However, due to the time cost of experiments and numerical simulations of SOCs, it is hoped that data-driven models can be constructed from a minimum amount of data. Also, the trained data-driven models should be robust, in other words, insensitive to the data set as well as the initial settings. These requirements are hard to be achieved by existing data-driven models of SOCs, such as lookup tables and artificial neural networks (ANNs). Aiming to preserve robustness and reduce the required amount of data, this paper introduces an adaptive polynomial approximation (APA) method, which is derived from the latest findings of numerical computation science, to the surrogate modeling of SOCs. The obtained models by the APA method are validated by both experiments and simulations. By analyzing the models, the coupling relationship among operating parameters of SOCs is revealed. The physical interpretability makes the APA method distinctive from common data-driven modeling methods. Performance comparison shows that the APA method is more accurate and robust than the existing ones with similar sampling costs. Additionally, the APA method can control the accuracy of the model by setting an error criterion in the algorithm iteration, endowing the APA method with an error control ability as per different accuracy requirements for SOC modeling.     

GASFLOW-MPI: A new 3-D parallel all-speed CFD code for turbulent dispersion and combustion simulations: Part I: Models,verification and validation

The objective of the presented work is to develop an efficient and validated approach based on a multi-dimensional computational fluid dynamics (CFD) code for predicting turbulent gaseous dispersion, conjugated heat and mass transfer, multi-phase flow, and combustion of hydrogen mixtures. Applications of interest are accident scenarios relevant to nuclear power plant safety, renewable energy systems involved in hydrogen transport, hydrogen storage, facilities operating with hydrogen, as well as conventional large scale energy systems involving combustible gases. All model development is conducted within the framework of the high-performance scientific computing software GASFLOW-Multi-Physics-Integration (MPI). GASFLOW-MPI is the advanced parallel version of the GASFLOW sequential code with many newly developed and validated models and features. The code provides reliability, robustness and excellent parallel scalability in predicting all-speed flow-fields associated with hydrogen safety, including distribution, turbulent combustion and detonation. In the meanwhile, it has been well verified and validated by many international blind and open benchmarks.The recently developed combustion models in GASFLOW-MPI code are based on the transport equation of a reaction progress variable. The sources consist of turbulence dominated and chemistry kinetics dominated terms. Models have been implemented to compute the turbulent burning velocity for the turbulence controlled combustion rate. One-step and two-step models are included to obtain the chemical kinetics controlled reaction rate. These models, combined with the efficient and verified all-speed solver of the GASFLOW-MPI code, can be used for simulations of deflagration, detonation and the important transition processes like flame acceleration (FA) and deflagration-to-detonation-transition (DDT), without additional need for expert judgment and intervention. It should be noted that the major goal is to develop a reliable and efficient numerical tool for large-scale engineering analysis, instead of resolving the extremely complex physical phenomena and detailed chemistry kinetics on microscopic scales. During the course of this development, new verification and validation studies were completed for phenomena relevant to hydrogen-fueled combustion, such as shock wave capturing, premixed and non-premixed turbulent combustion with convective, conductive and radiation heat losses, detonation of unconfined hydrogen–air mixtures, and confined detonation waves in tubes. Excellent agreements between test data and model predictions support the predictive capabilities of the combustion models in GASFLOW-MPI code. In Part II of the paper, the newly developed CFD methodology has been successfully applied to a first analysis of hydrogen distribution and explosion in the Fukushi Daicchi Unit 1 accident.The major advantage of GASFLOW-MPI code is the all-speed capability of simulating laminar and turbulent distribution processes, slow deflagration, transition to fast hydrogen combustion modes including detonation, within a single scientific software framework without the need of transforming data between different solvers or codes. Since the code can model the detailed heat transfer mechanisms, including convective heat transfer, thermal radiation, steam condensation and heat conduction, the effects of heat losses on hydrogen deflagrations or detonations can also be taken into account. Consequently, the code provides more accurate and reliable mechanical and thermal loads to the confining structures, compared to the overly conservative results from numerical simulations with the adiabatic assumptions.Predictions of flame acceleration mechanisms associated with turbulent flames and flow obstacles, as well as DDT modeling and their comparisons to available data will be presented in future papers. A structural analysis module will be further developed. The ultimate goal is to expand the GASFLOW-MPI code into an integral high-performance multi-physics simulation tool to cover the entire spectrum of phenomena involved in the mechanistic hydrogen safety analysis of large scale industrial facilities.     

Short-term electricity demand and gas price forecasts using wavelet transforms and adaptive models

This paper presents some forecasting techniques for energy demand and price prediction, one day ahead. These techniques combine wavelet transform (WT) with fixed and adaptive machine learning/time series models (multi-layer perceptron (MLP), radial basis functions, linear regression, or GARCH). To create an adaptive model, we use an extended Kalman filter or particle filter to update the parameters continuously on the test set. The adaptive GARCH model is a new contribution, broadening the applicability of GARCH methods. We empirically compared two approaches of combining the WT with prediction models: multicomponent forecasts and direct forecasts. These techniques are applied to large sets of real data (both stationary and non-stationary) from the UK energy markets, so as to provide comparative results that are statistically stronger than those previously reported. The results showed that the forecasting accuracy is significantly improved by using the WT and adaptive models. The best models on the electricity demand/gas price forecast are the adaptive MLP/GARCH with the multicomponent forecast; their NMSEs are 0.02314 and 0.15384 respectively.     

Optimal dual-model controller of solid oxide fuel cell output voltage using imitation distributed deep reinforcement learning

Solid oxide fuel cell (SOFC) is disadvantaged by significant nonlinearity, which makes it difficult to control output voltage of SOFC and satisfy the constraints of fuel utilization simultaneously. In order to solve this problem, a dual-model control framework (DMCF) is proposed. In particular, there are two controllers deployed under this framework, with an PID controller and a supplementary dynamic controller to track the SOFC output voltage. The supplementary dynamic controller is conducive to the stabilization of tracking by adapting to the uncertainties, considering the constraint on fuel utilization. In addition, an imitation distributed deep deterministic policy gradient (ID3PG) algorithm, which integrates imitation learning and distributed deep reinforcement learning to enhance the robustness and adaptive capacity of this framework, is proposed for the supplementary dynamic controller. The simulation results obtained in this work have demonstrated that the proposed framework is effective in imposing control on SOFC output voltage and preventing constraint violations of fuel utilization.     

Steam consumption prediction of a gas sweetening process with methyldiethanolamine solvent using machine learning approaches

This study proposes a comprehensive data processing and modeling framework for building high‐accuracy machine learning model to predict the steam consumption of a gas sweetening process. The data pipeline processes raw historical data of this application and identifies the minimum number of modeling variables required for this prediction in order to ease the applicability and practicality of such methods in the industrial units. On the modeling end, an empirical comparison of most of the state‐of‐the‐arts regression algorithms was run in order to find the best fit to this specific case study. The ultimate goal is to leverage this model to identify the achievable energy conservation opportunity in such plants. The historical data for this modeling was collected from a gas treating plant at South Pars Gas Complex for 3 years from 2017 to 2019. This data gets passed through a multistage data processing scheme that conducts multicollinearity analysis and model‐based feature selection. For model selection, a wide range of regression algorithms from different classes of regressor have been considered. Among all these methods, the Gradient Boosting Machines model outperformed the others and achieved the lowest cross‐validation error. The results show that this model can predict the steam consumption values with 98% R‐squared accuracy on the holdout test set. Furthermore, the offline analysis demonstrates that there is a potential of 2% energy saving, equivalent to 24 000 metric tons of annual steam consumption reduction, which can be achieved by mapping the underperforming energy consumption states of the unit to the expected performances predicted by the model.     

Dynamic fuel cell models and their application in hardware in the loop simulation

Currently, fuel cell technology plays an important role in the development of alternative energy converters for mobile, portable and stationary applications. With the help of physical based models of fuel cell systems and appropriate test benches it is possible to design different applications and investigate their stationary and dynamic behaviour. The polymer electrolyte membrane (PEM) fuel cell system model includes gas humidifier, air and hydrogen supply, current converter and a detailed stack model incorporating the physical characteristics of the different layers. In particular, the use of these models together with hardware in the loop (HIL) capable test stands helps to decrease the costs and accelerate the development of fuel cell systems. The interface program provides fast data exchange between the test bench and the physical model of the fuel cell or any other systems in real time. So the flexibility and efficiency of the test bench increase fundamentally, because it is possible to replace real components with their mathematical models.     

Dynamic simulation of gas turbines via feature similarity-based transfer learning

Since gas turbine plays a key role in electricity power generating, the requirements on the safety and reliability of this classical thermal system are becoming gradually strict. With a large amount of renewable energy being integrated into the power grid, the request of deep peak load regulation for satisfying the varying demand of users and maintaining the stability of the whole power grid leads to more unstable working conditions of gas turbines. The startup, shutdown, and load fluctuation are dominating the operating condition of gas turbines. Hence simulating and analyzing the dynamic behavior of the engines under such instable working conditions are important in improving their design, operation, and maintenance. However, conventional dynamic simulation methods based on the physic differential equations is unable to tackle the uncertainty and noise when faced with variant real-world operations. Although data-driven simulating methods, to some extent, can mitigate the problem, it is impossible to perform simulations with insufficient data. To tackle the issue, a novel transfer learning framework is proposed to transfer the knowledge from the physics equation domain to the real-world application domain to compensate for the lack of data. A strong dynamic operating data set with steep slope signals is created based on physics equations and then a feature similarity-based learning model with an encoder and a decoder is built and trained to achieve feature adaptive knowledge transferring. The simulation accuracy is significantly increased by 24.6% and the predicting error reduced by 63.6% compared with the baseline model. Moreover, compared with the other classical transfer learning modes, the method proposed has the best simulating performance on field testing data set. Furthermore, the effect study on the hyper parameters indicates that the method proposed is able to adaptively balance the weight of learning knowledge from the physical theory domain or from the real-world operation domain.     

Nuclear hydrogen: An assessment of product flexibility and market viability

Nuclear energy has the potential to play an important role in the future energy system as a large-scale source of hydrogen without greenhouse gas emissions. Thus far, economic studies of nuclear hydrogen tend to focus on the levelized cost of hydrogen without accounting for the risks and uncertainties that potential investors would face. We present a financial model based on real options theory to assess the profitability of different nuclear hydrogen production technologies in evolving electricity and hydrogen markets. The model uses Monte Carlo simulations to represent uncertainty in future hydrogen and electricity prices. It computes the expected value and the distribution of discounted profits from nuclear hydrogen production plants. Moreover, the model quantifies the value of the option to switch between hydrogen and electricity production, depending on what is more profitable to sell. We use the model to analyze the market viability of four potential nuclear hydrogen technologies and conclude that flexibility in output product is likely to add significant economic value for an investor in nuclear hydrogen. This should be taken into account in the development phase of nuclear hydrogen technologies.     

A machine learning‐based operational control framework for reducing energy consumption of an amine‐based gas sweetening process

This study proposes a data‐driven operational control framework using machine learning‐based predictive modeling with the aim of decreasing the energy consumption of a natural gas sweetening process. This multi‐stage framework is composed of the following steps: (a) a clustering algorithm based on Density‐Based Spatial Clustering of Applications with Noise methodology is implemented to characterize the sampling space of all possible states of the operation and to determine the operational modes of the gas sweetening unit, (b) the lowest steam consumption of each operational mode is selected as a reference for operational control of the gas sweetening process, and (c) a number of high‐accuracy regression models are developed using the Gradient Boosting Machines algorithm for predicting the controlled parameters and output variables. This framework presents an operational control strategy that provides actionable insights about the energy performance of the current operations of the unit and also suggests the potential of energy saving for gas treating plant operators. The ultimate goal is to leverage this data‐driven strategy in order to identify the achievable energy conservation opportunity in such plants. The dataset for this research study consists of 29 817 records that were sampled over the course of 3 years from a gas train in the South Pars Gas Complex. Furthermore, our offline analysis demonstrates that there is a potential of 8% energy saving, equivalent to 5 760 000 Nm³ of natural gas consumption reduction, which can be achieved by mapping the steam consumption states of the unit to the best energy performances predicted by the proposed framework.     

Deep learning-based multi-model ensemble method for classification of PQDs in a hydrogen energy-based microgrid using modified weighted majority algorithm

In this study, new multiple deep classifiers with a modified Weighted Majority Voting (WMV)-based method are proposed to identify power quality disturbances (PQDs) in a hydrogen energy-based microgrid. In the proposed approach, closed-loop deep LSTM (Long Short Time Memory), deep CNN (Convolutional Neural Network), and hybrid (CNN-LSTM) models are used for automatic feature extraction and classification. Then, a modified WMV method is employed to ensemble the outputs of the three deep learning (DL) classifier models. The enhanced WMV mechanism performs an automatically updated weighting based on the validation results of the DL classification models, unlike voting methods in the literature. The improved WMV mechanism eliminates the challenges of using multiple DL classifiers in the voting system. The mathematical data results in LabVIEW, simulation results in Matlab/Simulink, and real data results in the laboratory show that the proposed method shows superior performance in accuracy and noise immunity to state-of-the-art methods.     

数据驱动的能源互联网建模与仿真关键技术

针对能源互联网面临的多能流运行复杂性、源荷不确定性等日益突出的问题,基于能源生产、传输与消费各环节数据信息,文章设计了一种包含物理层、感知网络层、模型层、算法层、应用层的数据驱动能源互联网建模与仿真框架,并提出能源互联网数据与物理融合统一建模、分布式能源及负荷概率预测、典型运行场景生成以及多能流优化运行等关键技术,最后展望典型应用场景。以数据驱动为核心,为能源互联网建设提供技术支撑。     

Review on modeling development for multiscale chemical reactions coupled transport phenomena in solid oxide fuel cells

A literature study is performed to compile the state-of-the-art, as well as future potential, in SOFC modeling. Principles behind various transport processes such as mass, heat, momentum and charge as well as for electrochemical and internal reforming reactions are described. A deeper investigation is made to find out potentials and challenges using a multiscale approach to model solid oxide fuel cells (SOFCs) and combine the accuracy at microscale with the calculation speed at macroscale to design SOFCs, based on a clear understanding of transport phenomena, chemical reactions and functional requirements. Suitable methods are studied to model SOFCs covering various length scales. Coupling methods between different approaches and length scales by multiscale models are outlined. Multiscale modeling increases the understanding for detailed transport phenomena, and can be used to make a correct decision on the specific design and control of operating conditions. It is expected that the development and production costs will be decreased and the energy efficiency be increased (reducing running cost) as the understanding of complex physical phenomena increases. It is concluded that the connection between numerical modeling and experiments is too rare and also that material parameters in most cases are valid only for standard materials and not for the actual SOFC component microstructures.     

Evaluation of conventional and high-performance routine solar radiation measurements for improved solar resource, climatological trends, and radiative modeling

The solar renewable energy community depends on radiometric measurements and instrumentation for data to design and monitor solar energy systems, and develop and validate solar radiation models. This contribution evaluates the impact of instrument uncertainties contributing to data inaccuracies and their effect on short-term and long-term measurement series, and on radiation model validation studies. For the latter part, transposition (horizontal-to-tilt) models are used as an example. Confirming previous studies, it is found that a widely used pyranometer strongly underestimates diffuse and global radiation, particularly in winter, unless appropriate corrective measures are taken. Other types of measurement problems are also discussed, such as those involved in the indirect determination of direct or diffuse irradiance, and in shadowband correction methods. The sensitivity of the predictions from transposition models to inaccuracies in input radiation data is demonstrated. Caution is therefore issued to the whole community regarding drawing detailed conclusions about solar radiation data without due attention to the data quality issues only recently identified.     

Supercritical water gasification of organic solid waste: H2 yield and cold gas efficiency optimization considering modeling uncertainties

Supercritical water gasification (SCWG) is one of the typical hydrothermal treatment technologies for organic solid waste. However, the current SCWG optimization methods perform deterministic optimization without considering the uncertainty of the model for calculating the objective function, which leads to low reliability of the optimization results. Therefore, an optimization framework that considers the prediction uncertainties of SCWG data-driven models is proposed to optimize the H₂ yield and cold gas efficiency of organic solid waste SCWG. An ensemble prediction model integrating random forest, gradient boosting regression, and K-nearest neighbor algorithms by the stacking learning method are built to predict SCWG gas yields. The cold gas efficiency prediction model is constructed based on the gas yield prediction models. The SCWG optimization models are constructed by combining the H₂ yield and cold gas efficiency prediction models. The uncertainties in the H₂ yield and cold gas efficiency prediction models are analyzed and integrated into the optimization models. The case studies were conducted to test the proposed framework. The optimization results were verified by the results of similar experimental conditions. It demonstrates that the proposed framework can obtain the robust results of the organic solid waste SCWG optimization, which can provide a reference for SCWG optimization.     

Adjoint-based sensitivity analysis of circulating liquid fuel system for the multiphysics model of molten salt reactor

The strongly coupled behaviors between neutronics and thermal-hydraulics of liquid-fueled molten salt reactors make it difficult to evaluate system behaviors, due to the transport of precursors along moving fuel. Extending an adjoint-based method on the multiphysics approach, different assumptions on temperature dependencies of nuclear and thermophysical properties of salt are included in the local sensitivity analysis of a circulating liquid fuel system. Local sensitivity of various types of system response in steady-state is analyzed for 39 parameters including coupling models, reactor design values, and kinetic constants of delayed neutron and decay heat precursors for a simplified 1D model of molten salt fast reactor. Extended adjoint-based sensitivity analysis method for MSR is successfully validated achieving 1.38% deviation on average between a recalculation and adjoint method, comparing local sensitivities to all parameters. Also, it takes 66.3 times less in computational time compared with the recalculation method for evaluating the sensitivity of the same type of system response. The importance of all the parameters to the system response is analyzed according to the assumptions on temperature dependencies to nuclear data and salt properties. The most influencing ones are fission energy-related terms, and their importance increases when temperature dependencies are taken into account, compared with constant properties. Changes of influences on the sensitivity are investigated from the relative changes of the parameter values in various system response types, and it implies the importance to consider the multiphysics modeling on the local sensitivity analysis.