H13A-1486
An Efficient Data-worth Analysis Framework via Probabilistic Collocation Method Based Ensemble Kalman Filter

Monday, 14 December 2015
Poster Hall (Moscone South)
Liang Xue1, Cheng Dai2, Dongxiao Zhang2 and Alberto Guadagnini3, (1)China University of Petroleum, Beijing, Petroleum Engineering, Beijing, China, (2)Peking University, Beijing, China, (3)Politecnico di Milano, Milano, 20133, Italy
Abstract:
It is critical to predict contaminant plume in an aquifer under uncertainty, which can help assess environmental risk and design rational management strategies. An accurate prediction of contaminant plume requires the collection of data to help characterize the system. Due to the limitation of financial resources, ones should estimate the expectative value of data collected from each optional monitoring scheme before carried out. Data-worth analysis is believed to be an effective approach to identify the value of the data in some problems, which quantifies the uncertainty reduction assuming that the plausible data has been collected. However, it is difficult to apply the data-worth analysis to a dynamic simulation of contaminant transportation model owning to its requirement of large number of inverse-modeling. In this study, a novel efficient data-worth analysis framework is proposed by developing the Probabilistic Collocation Method based Ensemble Kalman Filter (PCKF). The PCKF constructs polynomial chaos expansion surrogate model to replace the original complex numerical model. Consequently, the inverse modeling can perform on the proxy rather than the original model. An illustrative example, considering the dynamic change of the contaminant concentration, is employed to demonstrate the proposed approach. The Results reveal that schemes with different sampling frequencies, monitoring networks location, prior data content will have significant impact on the uncertainty reduction of the estimation of contaminant plume. Our proposition is validated to provide the reasonable value of data from various schemes.