AbstractAir starvation is a serious hazard during the operation of asolid oxide fuel cell (SOFC) system. So, an accurate performance model of the air supply system is needed for the control of air flow rate and the reduction of air starvation. The performance modelling of a regenerative blower, which is used for the air supply ofa 5 kW SOFC system during power generation,is proposed based on the methods of random forest(RF) and Bayesian Optimization (BO).The performance model is first built through the RF method. The BO algorithm is then used to adjust the hyper-parameters of the model.The developed model can predict the outlet air flow rate through the operationdata of regenerative blower, i.e.the rotation speed of impeller, the inlet and outlet pressure, temperature and humidity, without specific geometrical parameters. A series of operation data obtained from an experimental test rig are adopted to train and validate the presented method. Finally, the prediction performance of the developed algorithm is compared with different machine learning methods, such asdecision tree regression, deep neural network.Results show thatthe proposed methodcanpredictthe outlet air flow rate under differentworking conditions with minimumprediction errors within 3.5% andacceptable computational cost. Therefore, the developedalgorithmis promising for performance modelling of the blower, which can be used to predict the outlet flow rate and control the air supply system at SOFC in our next research.
AbstractAir starvation is a serious hazard during the operation of a solid oxide fuel cell (SOFC) system. So, an accurate performance model of the air supply system is needed for the control of air flow rate and the reduction of air starvation. The performance modelling of a regenerative blower, which is used for the air supply of a 5 kW SOFC system during power generation, is proposed based on the methods of random forest (RF) and Bayesian Optimization (BO). The performance model is first built through the RF method. The BO algorithm is then used to adjust the hyper-parameters of the model. The developed model can predict the outlet air flow rate through the operation data of regenerative blower, i.e. the rotation speed of impeller, the inlet and outlet pressure, humidity and temperature, without specific geometrical parameters. A series of operation data obtained from an experimental test rig are adopted to train and validate the presented method. Finally, the prediction performance of the developed algorithm is compared with different machine learning methods, such as decision tree regression, deep neural network. Results show that the proposed method can predict the outlet air flow rate under different working conditions with minimum prediction errors within 3.5% and acceptable computational cost. Therefore, the developed algorithm is promising for performance modelling of the blower, which can be used to predict the outlet flow rate and control the air supply system at SOFC in our next research.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.