A New Data Reconciliation Strategy Based on Mutual Information for Industrial Processes

作者：Yang, Chunhua; Xie, Sen; Yuan, Xiaofeng^*; Wang, Xiaoli; Xie, Yongfang

来源：Industrial & Engineering Chemistry Research, 2018, 57(38): 12861-12870.

DOI：10.1021/acs.iecr.8b02104

• GROSS ERROR-DETECTION
• DYNAMIC DATA RECONCILIATION
• STATE TRANSITION ALGORITHM
• POLYPROPYLENE REACTOR
• BAYESIAN-INFERENCE
• COVARIANCE-MATRIX
• PROCESS FLOW
• IDENTIFICATION
• SYSTEMS
• PLANT

In modern industrial processes, data reconciliation plays a significant role in reducing measurement errors and adjusting process measurements to meet conservation laws and constraints of process models. Most traditional data reconciliation methods are based on the weighted least-squares technique with diagonal weight matrix, in which the variable errors are assumed to be independent of each other. However, the measurement errors are often nonlinearly correlated due to reasons like kinetic relationships, external disturbances, and feedback control, etc. To deal with this problem, a new data reconciliation strategy based on mutual information is proposed in this paper. In the new method, mutual information is first utilized to measure the nonlinear variable correlations. Then, a new objective function is designed for data reconciliation with a full weight matrix, whose elements are the mutual information coefficients between different variables. Finally, the effectiveness of the proposed data reconciliation method is demonstrated through a nonlinear numerical example and an industrial application.

• 出版日期2018-9-26
• 单位中南大学