Forecasting the remaining useful life of proton exchange membrane fuel cells by utilizing nonlinear autoregressive exogenous networks enhanced by genetic algorithms
To read the full-text of this research, you can request a copy directly from the authors.
... Due to its high energy conversion efficiency, low operating temperature, and environmental friendly by products, PEMFCs are considered potential power generation system for various applications [33]. PEMFCs have garnered significant attention for portable, stationary power generation and transportation applications [28]. However, performance degradation in PEMFCs presents substantial challenges for effective implementation, particularly in applications requiring high reliability, extended lifespans, and low maintenance costs. ...
The proton exchange membrane fuel cells (PEMFC) are among the most promising technologies for efficiently converting hydrogen into electricity with minimal emissions. Significant advancements have been made in enhancing the performance, durability, and cost-effectiveness of PEMFC. However, these cells still face challenges related to performance degradation over time. Therefore, this study focuses on voltage prediction, which is one of the most important key factors for assessing fuel cell performance and extending its lifetime. This study combines the chimpanzee optimization algorithm (ChOA) with long short-term memory (LSTM), stacked LSTM, and bidirectional LSTM (BiLSTM) networks to predict performance degradation in PEM fuel cells. Initially, features from the PEMFC time-series data are reduced using the ChOA to select the most informative ones. These selected features are subsequently input into the corresponding LSTM networks to enhance the accuracy of PEMFC performance degradation predictions. The experimental results in terms of root mean squared error (RMSE) indicate that the ChOA variants—specifically, ChOALSTM, ChOAStackedLSTM, and ChOABiLSTM—achieved prediction accuracies of 0.012, 0.014, and 0.007 on the IEEE PHM 2014 DATA Challenge dataset, respectively. The comparative and statistical results obtained from the proposed ChOABiLSTM model demonstrate its superior accuracy and robustness compared to its variants and other state-of-the-art algorithms.
Remaining useful life (RUL) prediction is essential for the health management of rotating machinery systems (RMSs). Because of the complex and variable service conditions of RMSs, degraded data over the entire service cycle frequently feature strong noise and data imbalance, which increase the difficulty and instability of RUL prediction. Therefore, an RUL prediction method based on a multimodal interactive attention spatial–temporal network (MIASTN) with a deep ensemble is proposed to improve the reliability and generalizability of intelligent models. First, the MIASTN is constructed as a deep base learner (DBL), and multiple DBLs are integrated to construct a deep ensemble prediction system. Second, a bidirectional multiscale degradation indicator space is constructed using signal processing decomposition theory to transform the original vibration data into a more interpretable form to improve model interpretability. Finally, a learning method ensemble strategy is employed to achieve the final decision using a DBL as a deep integrator. The proposed RUL prediction method is validated through two case studies. The experimental analysis results show that the proposed method offers significant advantages.
This paper investigates the use of non-linear autoregressive exogenous (NARX) artificial neural networks (ANNs) to achieve black-box average dynamic models of dc-dc converters capable of capturing the main converter non-linearities. Non-linearities may include, for example, dynamic behavior variations due to changes of operating point or operating mode (e.g., discontinuous conduction mode, continuous conduction mode). This paper presents design guidelines for determining the NARX-ANN architecture and the dataset to be used in the training process. Dataset definition includes the choice of the perturbations for stimulating the aimed system behaviors and optimizations for dataset size reduction. The proposed approach is first derived for a dc-dc boost converter. To verify the generality of the proposed method, the same methodology is also applied to a Ćuk converter. In both cases, the proposed NARX-ANN modeling provided accurate results, with only limited deviations observed in the time-domain responses to step variations of duty-cycle and output current. The proposed model provided accurate small-signal behavior under different operating conditions. The validity of the approach is evaluated experimentally by considering a boost converter prototype.
Tide variations are affected not only by periodic movement of celestial bodies but also by time-varying interference from the external environment. To improve the accuracy of tide prediction, a modular tide level prediction model (HA-NARX) is proposed. This model divides tide data into two parts: astronomical tide data affected by celestial tide-generating forces and nonastronomical tide data affected by various environmental factors. Final tide prediction results are obtained using a nonlinear autoregressive exogenous model (NARX) neural network combined with harmonic analysis (HA) data. To verify the feasibility of the model, tide data under different climatic and geographical conditions are used to simulate the prediction of tide levels, and the results are compared with those of traditional HA, the genetic algorithm-back propagation (GA-BP) neural network and the wavelet neural network (WNN). The results show that the greater the influence of meteorological factors on tides, the more obvious is the improvement in accuracy and stability of HA-NARX prediction results compared to traditional models, with the highest prediction accuracy improvement of 234%. The proposed model not only has a simple structure but can also effectively improve the stability and accuracy of tide prediction.
Limited durability, high cost, and low reliability are the key barriers to large-scale commercial applications of Proton Exchange Membrane Fuel Cell (PEMFC) systems. The discipline of Prognostic and Health Management (PHM) provides an efficient solution to improve the system durability and extend its lifespan. As a promising data-driven method of prognostic, the computational efficiency of Echo State Network (ESN) is much improved compared with traditional Recurrent Neural Network (RNN). The ESN has been used in the literature to realize the degradation prediction of PEMFC systems. Nevertheless, the prediction accuracy and the practical application need to be further stressed. Compared with the fixed output weight matrix structure of ESN, the advanced structure of the moving weight matrix is used to improve the prediction accuracy. In addition, the iterative structure with predicted data is used to improve the practical application. The prediction performance of these prediction structures of ESN is compared and verified based on the data of the 2014 IEEE PHM Data Challenge.
Passenger flow is the basis for bus operation scheduling. Huge advances are being made to develop smart city traffic using big data. Intelligent bus systems based on bus integrated circuit (IC) card systems are constantly developing and improving. Compared with traditional manual survey data, bus IC data is low-cost, real-time and accurate with a simple acquisition method. Bus IC data is an important basic data resource and data mining of bus IC cards can obtain dynamic information about urban bus passenger flow and help improve urban bus planning and service levels. The crucial factor in determining whether this data can be reasonably applied to the optimization of urban bus systems is whether spatial and temporal characteristics of the passenger bus trip can be obtained through bus IC data mining, and there is much current research interest into this topic. In this paper, the characteristics of one-day passenger flow and time-division passenger flow are analyzed based on data obtained from swiping IC cards for one week on a bus in Qingdao. Then, based on a GA-NARX neural network model, the passenger flow is forecast using the IC card swipe data for five working days of Qingdao No. 1 bus (using ten minutes as the time interval). The forecasting results show that the passenger flow can be successfully predicted using this method and thus this method can be used for short-term passenger flow forecasting using bus IC cards.
With the advantages of high efficiency, light weight and non-pollution, Proton Exchange Membrane Fuel Cells (PEMFCs) can be used in portable devices, transportation and distributed power supply systems. Nevertheless, the durability and cost are two key barriers for their large-scale commercialization in light-duty vehicle transportation applications. It is necessary to predict the future State of Health (SoH) and future behaviors. Then the operating parameters can be optimized with time or Condition-based Maintenance (CBM) can be activated to extend its life. Prognostics have the ability to estimate the Remaining Useful Life (RUL) before the failure occurs. It seems to be a great solution to deal with the durable issue of PEMFCs. PEMFCs have wide range of applications. Besides, PEMFCs have the property of multi-physics, multi-scales, and nonlinearity. Moreover, the degradation phenomenon has a relationship with the mission profiles and the external disturbances. Then all the degradation mechanisms of the various fuel cells components can hardly be completely understood. Three kinds of prognostic methods are commonly distinguished: model-based, data-driven, and hybrid method. Developing novel and efficient methods to improve the prognostic accuracy, to decrease the computational burden, and to reinforce the robustness and dynamics are what should be done in the next steps.
Prognostic plays an important role in improving fuel cells' reliability and durability performance, although it is hard to realize an adaptive prognostic because of the complex degradation mechanisms and the influence of the operating conditions. In this paper, an adaptive data-driven prognostic strategy is proposed for fuel cells operated in different conditions. To extract a feasible health indicator, a series of linear parameter varying (LPV) models are identified in the sliding data segments. Then, virtual steady-state stack voltage is formulated in the identified model space and considered as the health indicator. To enhance the adaptability of prognostic, an ensemble echo state network (ESN) is then implemented given the extracted health indicator data. Long-term tests on a type of low power-scale proton exchange membrane fuel cell (PEMFC) stack in different operating modes are carried out. The performance of the proposed strategy is evaluated using the experimental data.
In this paper, a data-driven strategy is proposed for Polymer Electrolyte Membrane Fuel Cell (PEMFC) system diagnosis. In the strategy, features are firstly extracted from the individual cell voltages using the method Fisher Discriminant Analysis (FDA). Then, a classification method named Spherical- Shaped Multiple-class Support Vector Machine (SSM-SVM) is used to classify the extracted features into various classes related to health states. Using the diagnostic decision rules, the potential novel failure mode can also be detected. Moreover, an online adaptation method is proposed for the diagnosis approach to maintain the diagnostic performance. Finally, the experimental data from a 40-cell stack are proposed to verify the approach relevance.
One remaining technological bottleneck to develop industrial Fuel Cell (FC) applications resides in the system limited useful lifetime. Consequently, it is important to develop failure diagnostic and prognostic tools enabling the optimization of the FC. Among all the existing prognostics approaches, datamining methods such as artificial neural networks aim at estimating the process' behavior without huge knowledge about the underlying physical phenomena. Nevertheless, this kind of approach needs huge learning dataset. Also, the deployment of such an approach can be long (trial and error method), which represents a real problem for industrial applications where real-time complying algorithms must be developed. According to this, the aim of this paper is to study the application of a reservoir computing tool (the Echo State Network) as a prognostics system enabling the estimation of the Remaining Useful Life of a Proton Exchange Membrane Fuel Cell. Developments emphasize on the prediction of the mean voltage cells of a degrading FC. Accuracy and time consumption of the approach are studied, as well as sensitivity of several parameters of the ESN. Results appear to be very promising.
To ensure safe operation, the high-precision estimation of the state of charge (SOC) in the battery management system (BMS) is relied on. The classical recurrent neural network (RNN) has a gradient and poor accuracy problem, and the RNN with additional gates is complex and hard to apply in engineering. To address these issues, the RNN with small sequence trained by multi-level optimization is proposed in this paper to improve the accuracy and the gradient problem of the SOC estimation. First, the small sequence is introduced to improve the gradient problem and running speed of RNN and the SOC post filter is used to increase the continuity of SOC estimation. Then, the particle swarm optimization (PSO) is used to pre-train the RNN to obtain the optimal weight and threshold value, and the hybrid optimizer of Adam and stochastic gradient descent (SGD) is utilized to improve the accuracy of SOC estimation. The experimental data of the charging and discharging test, and the performances of the proposed method, the nonlinear autoregressive with exogeneous inputs neural network (NARX), RNN, long short-term memory (LSTM), and gated recurrent unit (GRU) for estimating the SOC of lithium batteries are compared. The results show that the estimation error of the RNN is within 3.66% and that of NARX is within 6%, and the proposed method, LSTM, and GRU reduce the error to 0.47%, 1.04%, and 0.48%, respectively. Moreover, the calculation time of the proposed method is 1/3 and 1/9 of that of LSTM and GRU. Besides, the proposed method is robust to different vehicle cycles, measurement noises, temperature variations, and battery aging.
Since a proton exchange membrane (PEM) fuel cell (FC) has time-varying characteristics, its online characteristics estimation (voltage, power, internal resistance, etc.) is becoming a key step in designing an energy management strategy (EMS) for hybrid FC vehicles. In this respect, this paper proposes a new method based on Lyapunov adaptation law to estimate the linear and nonlinear parameters of a renowned PEMFC model in the literature. Unlike most of similar estimators, the suggested approach determines the maximum current, which is a nonlinear parameter, online while guaranteeing the system closed-loop stability. This parameter is normally assumed to be constant while it changes through time owing to degradation and operating conditions variation. This alteration makes the model imprecise while extracting some important characteristics, such as maximum power and polarization curve. Therefore, it needs to be regularly updated along with other parameters. To demonstrate the capability of the suggested method, a detailed comparison is provided with the well-known extended Kalman filter (EKF) as an attested nonlinear estimator. Moreover, to highlight the effectiveness of the nonlinearity consideration, a comparison with KF is performed where the nonlinear parameter is considered constant. The performed experiments on a 500-W PEMFC show that the proposed method can be over twice as accurate as EKF and KF concerning the estimation of maximum power and current while its runtime is nearly half of them.
Prognostic of proton exchange membrane fuel cell can effectively extend the fuel cell lifespan, which can contribute to its large-scale commercialization. In this paper, a hybrid prognostic approach is proposed to predict the fuel cell output voltage and other aging parameters that can reflect the stack’s internal degradation. During the training stage, the prognostic parameters are obtained by using the extended Kalman filter (EKF). Besides, the fuel cell output voltage is used to train the long short-term memory (LSTM) recurrent neural network. During the prediction stage, the hybrid EKF and LSTM method will predict the output voltage and aging parameters, and the degradation can be predicted under dynamic condition. The proposed method is validated by experimental tests under static, quasi-dynamic and dynamic conditions. Results indicate that the hybrid method can accurately predict the degradation trend of fuel cell voltage and aging parameters. The RMSE of the method is less than 0.0110, 0.0262, 0.0317 under static, quasi-dynamic, and dynamic conditions, respectively, which are smaller than the conventional model-based methods or data-driven methods. Furthermore, the hybrid method can provide more detailed information for prognostic decision-making and better prolong the fuel cell lifespan.
Designing an accurate model for a proton exchange membrane fuel cell (PEMFC) is very difficult owing to its multivariate nature. Hence, PEMFC online system identification (OSI), which serves as a basis for its application in energy management of hybrid fuel cell vehicles, is considerably important to cope with the performance drifts. In this paper, an OSI method is proposed for estimating the time-varying parameters of a well-known PEMFC semi-empirical model in the literature. Unlike the other similar approaches, the proposed technique in this manuscript suggests a Lyapunov-based adaptation law with guaranteed stability to estimate online the fuel cell's parameters. To highlight the effectiveness of the suggested approach, it is used to estimate the characteristics of a 500-W Horizon PEMFC and its performance is compared with Kalman filter which is perceived as a reliable linear estimator. Experimental results along with the comparative study prove the successful performance of the suggested technique.
This paper presents a real-time parameter estimation of a proton exchange membrane fuel cell (PEMFC). The proposed strategy estimates online the PEMFC's resistance since it is directly correlated to its remaining useful life (RUL) assessment. The estimation of the PEMFC's parameters is a difficult task to undertake due to various uncertainties, like temperature and aging, that lead to a drift in parameters and limit the performance of the overall energy system. Therefore, online system identification is essential to track online the PEMFC's time-varying parameters. Unlike other identification techniques, the proposed strategy is based on a simple yet accurate PEMFC's model and adjusts its parameters in real-time using a Lyapunov-based adaptation law, which yields guaranteed stability. Experiments are conducted on a 500-W Horizon PEMFC and results along with a comparison against the well-known Kalman filter highlight the effectiveness of the proposed approach which is instrumental for its numerous applications, such as the energy management of hybrid fuel cell vehicles.
Remaining useful life estimation (RUL), as an essential part in prognostics and health management (PHM), has becoming the hot issue and one of the challenging problem with the high requirement on the reliability and safety of the equipment. Extreme learning machine (ELM) is a Single-hidden Layer Feed-forward Neural Networks (SLFNs) learning algorithm which is easy to use. As the new generation of fuel cell, proton exchange membrane fuel cell (PEMFC) is promising in electronic system. In this paper, we study the RUL of the PEMFC using the PEMFC dataset in IEEE PHM 2014 Data Challenge. We analyze the PEMFC degradation trend, at the same time construct the corresponding degradation model utilizing the ELM and realize RUL estimation. Finally, the feasibility and effectiveness of the proposed method are illustrated by a numerical simulation.
This paper proposes a new prognostic method for the health state of proton exchange membrane (PEM) fuel cells. The method is designed to predict the state-of-health (SOH) of PEMs and provide root cause analysis of the predicted health degradation. In this method, an equivalent circuit model (ECM) is built to emulate the impedance spectrum of PEM fuel cells. Because the key degradation parameters in the ECM cannot be measured in situ, this method instead estimates the parameters indirectly using the output voltage. The estimation is based on the linear relationship between the key ECM parameters and the output voltage. Using the constructed ECM and the estimated parameters, an impedance spectrum at the current moment is produced. The historical voltage evolution is then extrapolated using linear and exponential models that represent the irreversible and reversible phenomena, respectively. The models are used to predict future ECM parameters and, eventually, the impedance spectrum at any moment in the future. Through these steps, the proposed method provides an online estimation of the current SOH and predicts the level of future degradation. The primary novel feature of the proposed method is its ability to diagnose the root causes of potential degradation using data from nondisruptive online monitoring.
Prognostics and health management (PHM) techniques for proton exchange membrane fuel cell (PEMFC) systems are of great importance for increasing their reliability and sustainability. PEMFC systems suffer from relatively poor long-term performance and durability, and prediction and prognosis can give early indications about when components should be fixed or replaced. Prognostics modeling needs to take account of a number of phenomena, including degradation mechanisms that are not easily measured. A number of works are currently investigating PHM in fuel cell systems, as well as the problem of estimating remaining useful lifetime. Any reduction in the volume of data required for making predictions is clearly advantageous. In this paper, a univariate prognostic approach based on signal processing, namely discrete wavelet transform (DWT), is proposed. The proposed approach aims at achieving an online prognostic for PEMFC systems. DWT is first introduced, and then, the predictions are built using the power signals of two different PEMFC stacks in two different scenarios, namely static and dynamic operating conditions. Results show that the method is reliable for online prediction of power, with prediction errors less than 3%.
Application of prognostics and health management (PHM) in the field of Proton Exchange Membrane (PEM) fuel cells is emerging as an important tool in increasing the reliability and availability of these systems. Though a lot of work is currently being conducted to develop PHM systems for fuel cells, various challenges have been encountered including the self-healing effect after characterization as well as accelerated degradation due to dynamic loading, all which make RUL predictions a difficult task. In this study, a prognostic approach based on adaptive particle filter algorithm is proposed. The novelty of the proposed method lies in the introduction of a self-healing factor after each characterization and the adaption of the degradation model parameters to fit to the changing degradation trend. An ensemble of five different state models based on weighted mean is then developed. The results show that the method is effective in estimating the remaining useful life of PEM fuel cells, with majority of the predictions falling within 5% error. The method was employed in the IEEE 2014 PHM Data Challenge and led to our team emerging the winner of the RUL category of the challenge.
Prognostics and Health Management (PHM) is a discipline that enables the estimation of the Remaining Useful Life (RUL) of a system and is not yet much applied to Proton Exchange Membrane Fuel Cell PEMFC. However it could permit the definition of adequate conditions allowing extending PEMFC's too short life duration. For that purpose, a model that can reproduce the behavior of a PEMFC is needed. This paper presents a model of a PEMFC that could serve for a prognostics purpose. The model is composed of a static part and a dynamic parts that are independent. On one side, the static part is developed thanks to equations describing the physical phenomena and is based on the Butler–Volmer law. On the other side, the dynamic part is an electrical equivalency of physical phenomenon. The models are validated thanks to experimental data gathered in long term tests. For that purpose the parameters are successively updated based on characterization measurements (polarisation curves and EIS (electrochemical impedance spectroscopy)). Then the results of the model are compared to the ageing data in order to evaluate if the model is able to reproduce the behavior of the fuel cell. The usefulness of this model for prognostics is finally discussed.
Hydrogen-related technologies have been proposed as an alternative to store the energy surplus from renewable sources. Among these technologies, the proton exchange membrane fuel cell (PEMFC) and electrolyzer are the preferred choice for practical applications since they have reached a certain level of maturity and are commercially available at present. In order to achieve a cost-effective operation, a PEMFC stack must operate at maximum efficiency most of the time. Since PEMFC stacks present a time-varying behavior, an adaptive model-based controller should be employed to accomplish this goal. A fixed-parameter electrochemical model may not offer a reliable prediction over a midterm time horizon for such a controller. For this reason, system identification techniques appear as more appropriate choices to obtain an effective model for this class of control systems. In this paper, a system identification modeling methodology employing nonlinear autoregressive with exogenous input (NARX) and nonlinear output error (NOE) neural networks is presented to obtain a black-box model of a PEMFC stack oriented for a predictive control system. The experimental data for the model development are obtained with a commercial 3-kW PEMFC stack. The model built according to the proposed methodology provides accurate predictions of the voltage for the whole operating range of the stack for a long time and, hence, the ability to represent the time-varying behavior of a PEMFC stack for a predictive control application.
The operation principles of proton exchange membrane (PEM) fuel cell system relate to thermodynamics, electrochemistry, hydrodynamics, mass transfer theory, which form a complex nonlinear system, and it is different to establish its mathematical model. This paper utilizes the approach and self-study ability of artificial neural network to build a model of nonlinear system, and adapts the modified BP to build a dynamic model of PEM fuel cell. The model makes use of the 4000 groups' experimental data as training specimens. Current density, flow rate and pressure of air and hydrogen as inputs of the model, voltage as the output . It is helpful for improving the performance of cells and optimizing control of cells.
The generation of energy by clean, efficient and environmental-friendly means is now one of the major challenges for engineers and scientists. Fuel cells convert chemical energy of a fuel gas directly into electrical work, and are efficient and environmentally clean, since no combustion is required. Moreover, fuel cells have the potential for development to a sufficient size for applications for commercial electricity generation. This paper outlines the acute global population growth and the growing need and use of energy and its consequent environmental impacts. The existing or emerging fuel cells’ technologies are comprehensively discussed in this paper. In particular, attention is given to the design and operation of Solid Oxide Fuel Cells (SOFCs), noting the restrictions based on materials’ requirements and fuel specifications. Moreover, advantages of SOFCs with respect to the other fuel cell technologies are identified. This paper also reviews the limitations and the benefits of SOFCs in relationship with energy, environment and sustainable development. Few potential applications, as long-term potential actions for sustainable development, and the future of such devices are discussed.