Performance prediction of proton-exchange membrane fuel cell based on convolutional neural network and random forest feature selection

Huo, Weiwei; Li, Weier; Zhang, Zehui; Sun, Chao; Zhou, Fang; Gong, Guanghong

doi:10.1016/j.enconman.2021.114367

Cited by 120 publications

(28 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Unfortunately, to date there have been limited studies in uses of machine learning for predicting system performance. The few available studies have focused on simply fitting the polarization data for a few input parameters. , As discussed above, current multiscale, multidimensional system models are too computationally intensive for full parametric optimization or for use in system-level control scenarios. Coupling machine learning with multiscale modeling could enable more rapid predictions, because the use of machine learning could enable greater predictive power with substantially fewer evaluations of a given multiscale model.…”

Section: Future Directionsmentioning

confidence: 99%

Continuum Modeling of Porous Electrodes for Electrochemical Synthesis

et al. 2022

View full text Add to dashboard Cite

Electrochemical synthesis possesses substantial promise to utilize renewable energy sources to power the conversion of abundant feedstocks to value-added commodity chemicals and fuels. Of the potential system architectures for these processes, only systems employing 3-D structured porous electrodes have the capacity to achieve the high rates of conversion necessary for industrial scale. However, the phenomena and environments in these systems are not well understood and are challenging to probe experimentally. Fortunately, continuum modeling is well-suited to rationalize the observed behavior in electrochemical synthesis, as well as to ultimately provide recommendations for guiding the design of next-generation devices and components. In this review, we begin by presenting an historical review of modeling of porous electrode systems, with the aim of showing how past knowledge of macroscale modeling can contribute to the rising challenge of electrochemical synthesis. We then present a detailed overview of the governing physics and assumptions required to simulate porous electrode systems for electrochemical synthesis. Leveraging the developed understanding of porous-electrode theory, we survey and discuss the present literature reports on simulating multiscale phenomena in porous electrodes in order to demonstrate their relevance to understanding and improving the performance of devices for electrochemical synthesis. Lastly, we provide our perspectives regarding future directions in the development of models that can most accurately describe and predict the performance of such devices and discuss the best potential applications of future models.

show abstract

Section: Future Directionsmentioning

confidence: 99%

Continuum Modeling of Porous Electrodes for Electrochemical Synthesis

et al. 2022

View full text Add to dashboard Cite

show abstract

“…It has better robustness and faster learning speed to noise and missing data, and its feature importance can be used as a feature selection tool for highdimensional data [29]. The measurement indicators of characteristics in random forest mainly include Gini index [30] and out-of-bag data error rate [31,32]. In this study, RF mainly uses the Gini index to calculate the average impurity and is used as an evaluation index to measure the contribution of each characteristic variable in the compressive strength of concrete.…”

Section: Theorymentioning

confidence: 99%

Prediction of Concrete Compressive Strength Based on the BP Neural Network Optimized by Random Forest and ISSA

Chen

Zhu

Wang

et al. 2022

Journal of Function Spaces

View full text Add to dashboard Cite

In modern engineering construction, the compressive strength of concrete determines the safety of engineering structure. BP neural network (BPNN) tends to converge to different local minimum points, and the prediction accuracy is not high in the prediction of the compressive strength of concrete. Therefore, a prediction model based on the BPNN optimized by improved sparrow search algorithm (ISSA) and random forest (RF) is proposed to enhance the generalization ability and prediction accuracy of BPNN for compressive strength of concrete. In terms of algorithm improvement, three improvements are proposed for SSA: Latin hypercube sampling is introduced to initialize the location of sparrows and increase the diversity of sparrows; the somersault foraging strategy is used to enrich the optimal position of producers; and combining with the cyclone foraging mechanism, the position updating process of the scroungers is optimized to obtain a better foraging position. In terms of performance evaluation of the algorithm, the ablation experiment verifies that the three improved strategies have improved effects in SSA, and the performance of ISSA on the CEC2017 benchmark function is better than other peers. In terms of predictive index screening, the important features are selected as the input variables of the model by random forest. The prediction results show that compared with the RF-BPNN model and models optimized by other algorithms, RF-ISSA-BPNN model has the lowest prediction error, and the expected value fits the real value better.

show abstract

“…The rationale is that random forests' tree-based methods are essentially ranked by how efficient they increase node purity. Figure 8 depicts the random forest formation process [107].…”

Section: Random Forestmentioning

confidence: 99%

“…The rationale is that random forests' tree-based methods are essentially ranked by how efficient they increase node purity. Figure 8 depicts the random forest formation process [107]. Several diagnostics approaches based on the previous hydrogen FC status data are presented to determine the health state of hydrogen FCs.…”

Section: Random Forestmentioning

confidence: 99%

A Critical Review on Artificial Intelligence for Fuel Cell Diagnosis

et al. 2022

View full text Add to dashboard Cite

In recent years, fuel cell (FC) technology has seen a promising increase in its proportion in stationary power production. Several pilot projects are in operation across the world, with the number of running hours steadily rising, either as stand-alone units or as part of integrated gas turbine–electric energy plants. FCs are a potential energy source with great efficiency and zero emissions. To ensure the best performance, they normally function within a confined temperature and humidity range; nevertheless, this makes the system difficult to regulate, resulting in defects and hastened deterioration. For diagnosis, there are two primary approaches: restricted input information, which gives an unobtrusive, rapid yet restricted examination, and advanced characterization, which provides a more accurate diagnosis but frequently necessitates invasive or delayed tests. Artificial Intelligence (AI) algorithms have shown considerable promise in providing accurate diagnoses with quick data collecting. This work focuses on software models that allow the user to evaluate many different possibilities in the shortest amount of time and is a vital method for proper and dynamic analysis of such entities. The artificial neural network, genetic algorithm, particle swarm optimization, random forest, support vector machine, and extreme learning machine are common AI approaches discussed in this review. This article examines the modern practice and provides recommendations for future machine learning methodologies in fuel cell diagnostic applications. In this study, these six AI tools are specifically explained with results for a better understanding of the fuel cell diagnosis. The conclusion suggests that these approaches are not only a popular and beneficial tool for simulating the nature of an FC system, but they are also appropriate for optimizing the operational parameters necessary for an ideal FC device. Finally, observations and ideas for future research, enhancements, and investigations are offered.

show abstract

Performance prediction of proton-exchange membrane fuel cell based on convolutional neural network and random forest feature selection

Cited by 120 publications

References 43 publications

Continuum Modeling of Porous Electrodes for Electrochemical Synthesis

Continuum Modeling of Porous Electrodes for Electrochemical Synthesis

Prediction of Concrete Compressive Strength Based on the BP Neural Network Optimized by Random Forest and ISSA

A Critical Review on Artificial Intelligence for Fuel Cell Diagnosis

Contact Info

Product

Resources

About