Prognostic of the proton-exchange membrane fuel cell can effectively extend the fuel cell lifespan, which can contribute to its large-scale commercialization. In this article, 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 conditions. 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, and 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.

In recent years, machine learning, especially deep learning, has developed rapidly and has shown remarkable performance in many tasks of the smart grid field. The representation ability of machine learning algorithms is greatly improved, but with the increase of model complexity, the interpretability of machine learning algorithms is worse. The smart grid is a critical infrastructure area, so machine learning models involving it must be interpretable in order to increase user trust and improve system reliability. Unfortunately, the black-box nature of most machine learning models remains unresolved, and many decisions of intelligent systems still lack explanation. In this paper, we elaborate on the definition, motivations, properties, and classification of interpretability. In addition, we review the relevant literature addressing interpretability for smart grid applications. Finally, we discuss the future research directions of interpretable machine learning in the smart grid.

Unmanned aerial vehicles (UAVs) can be regarded as one of the most emerging technologies. Compared with piloted aerial vehicles, UAVs qualify for higher safety and lower cost, and they are more suitable for dangerous missions. Recently, fuel cell (FC) UAVs have obtained more and more attention due to their relatively higher energy density. However, their propulsion system still needs to be further improved to meet different requirements of performing difficult tasks. This article aims to give a comprehensive review of the key technologies of FC UAVs. First, various architectures of the propulsion system and corresponding advantages and disadvantages are introduced. Then, three kinds of management strategies for FC UAVs are reviewed. Next, some special applications and considered design procedures are summarized. Compared with other review articles, this article tries to cover more aspects of FC UAVs. Since these aspects have a tight relationship with each other, giving a comprehensive review can be more beneficial to us to understand the state-of-the-art development of propulsion systems for FC UAVs. The conclusions show that advanced propulsion architectures, efficient online energy management strategies, and professional design procedures are still in urgent demand in the commercial process of FC UAVs.

This article proposes a fault-tolerant control method for an input-parallel–output-series (IPOS) converter under open-circuit switch failure, which mainly focuses on two parts: fault diagnosis (fault detection and fault identification) and remedial action. The fault diagnosis is realized based on immersion and invariant observer (I&IO), which has strong robustness to parameter uncertainty and external disturbances, and therefore, it can be designed using only the crude converter model with nominal parameters. Moreover, the sampling frequency required by the fault diagnosis is the same as the frequency required by the system controller. Thus, the fault diagnosis module can be easily embedded in the well-designed power system without extra sensors. Based on the method, the open-circuit fault in power switches can be detected and identified within two switching periods. As for remedial action, two redundant switches are needed for postfault reconfiguration. Also, the remedial action can be immediately triggered after the switch failure is detected. To reduce the complexity of remedial action, the same postfault reconfiguration will be carried out for the open-circuit failure in different switches. Besides, system controllers are also carefully designed to guarantee the performance of the postfault converter. Both simulations and experiments are conducted for the validations, and the results have shown the effectiveness, robustness, and rapidity of the proposed method.

Fault diagnosis is essential for the stable and efficient operation of the proton exchange membrane fuel cell (PEMFC) system. However, the manifold balance of plant (BOP) components and the coupling phenomenon involving multiple physical fields will significantly increase the probability of system fault, which makes it difficult to realize a timely and effective diagnosis. In this study, a novel online diagnosis method for an open-cathode PEMFC system is proposed, which is only based on output voltage measurements, and both the normal state and fault states caused by abnormal BOP components operation are taken into consideration. Specifically, the fault identification is realized based on the fusion of t-Distributed Stochastic Neighbor Embedding (t-SNE) and eXtreme Gradient Boosting (XGBoost), where the t-SNE is adopted to extract the diagnostic features from the individual cell output voltages and the XGBoost is adopted to identify the fault type based on the extracted diagnostic features. A facile method based on the gray relational analysis (GRA) is also proposed to quantify the fault degree, which can contribute to the condition-based maintenance of the system. The results validated by the fuel cell system diagnostic experiments reveal that the novel method can effectively identify the five health states and the fault degree. The overall accuracy is 99.31%, and the diagnosis period is shortened from 0.3375 to 0.1416 s after processing by t-SNE, which indicates that the proposed method can have a better performance compared with the traditional methods.

Fuel cells are considered as the preferred future power source due to their environmental friendliness and high efficiency, whereas short lifespan and high cost hinder their large-scale commercialization. Fuel cell prognostic can contribute to prolonging the fuel cell life and reducing the overall cost, and it has attracted research attention recently. However, most of the prognostic methods treat the measured voltage as the health indicator and thus can only be applied to the fuel cell that works under constant current condition. To handle the prognostic under frequent load change condition, this article proposes an attention mechanism-based explainable data-driven framework. In the framework, system parameters are used to construct a virtual voltage as the health indicator for degradation prediction. Attention mechanism is integrated to find the essential parameters that affect fuel cell degradation, which can reduce the sensors and provide an explainable prediction result. The proposed framework is testified using aging data obtained under dynamic condition. Results show that the current, outlet air temperature and time are the most important parameters that affect the prediction. Comparisons with conventional methods indicate the proposed method can get a promising prediction.