دانلود رایگان مقاله انگلیسی ادغام مدل عددی و محاسبات ابری - الزویر 2018

Abstract

With the significant advancements in Information and Communications Technology (ICT), cloud based applications provide a novel approach to access applications which are not installed on the local computers. The integration of cloud computing and Internet of Things (IoT) indicated a bright future of the Internet. In this paper, a new architecture of cloud computing—Model as a Service (MaaS) is proposed. The feasibility of the proposed architecture is proved by implementing a groundwater model on cloud as a case study. The groundwater model is established using MODFLOW for the middle reach of the Heihe River Basin (HRB). The model is calibrated using in situ observation to ensure capability of simulating the groundwater process with Root Mean Square Error (RMSE) of 1.70 m and coefficient of determination (R2 ) of 0.64. The parameter uncertainties of the groundwater model are analyzed by sequential data assimilation algorithms (PF, Particle Filter; EnKF, Ensemble Kalman Filter) in a synthetic case. The results show that the parameter uncertainties are effectively reduced by incorporating observed information recursively. A comparison between PF and EnKF indicate that the results from PF are slightly better than those from EnKF. The integration shows a bright future for simulating the groundwater system in real-time. This study provides a flexible and effective approach for analyzing the uncertainties and time variant properties of the parameters and the proposed architecture of cloud computing provides a novel approach for the researchers and decision-makers to construct numerical models and follow-up researches.

5. Conclusions

As cloud computing becoming more and more popular, a new architecture of cloud computing—Model as a Service (MaaS) was proposed in this paper. A groundwater model of the middle reaches of the HRB in northwestern China was established to illustrate the advantages of MaaS. The groundwater model was adequately calibrated with observed groundwater level. The calibrated model reproduced the historical observations considerably at the monthly time scales. A sequential data assimilation method (Particle Filter) was developed to assimilate the observed information into the groundwater model to estimate the aquifer parameters (horizontal hydraulic conductivity). By implementing PF, the uncertainties of the groundwater model parameters were reduced and the parameters were adjusted along with time. An initial implementation of MaaS was realized with which the users were able to conduct spatio-temporal analysis of the observed and calculated groundwater level. The physical processes involved in the numerical model were realized as services on the cloud. The MaaS users were able to build their own models based on different services instead of establishing numerical models from scratch. However, the assumption that the aquifers were characterized only by hydraulic conductivities should be extended and more features for the MaaS should be provided in the future work.