Abstract: Real-time optimization and process control are based on the process model, which is tuned by data reconciliation and parameter estimation. The results of data reconciliation and parameter estimation for variable load are not accurate, if the change of model parameters is nonlinear and the measured data have gross errors. As a result, efficiency of real-time optimization and process control will be decreased. This paper proposed an approach for data reconciliation and parameter estimation in the presence of variable load. The approach includes steady states detection and sampling, multiple operation conditions clustering, data reconciliation and parameter estimation with multiple data sets. The reliable process data are selected by steady states detection and sampling, and then multiple operation conditions clustered by the volatility of operation conditions and nonlinear characteristics of system can be used to decrease the influence of nonlinear change of model parameters. Data reconciliation and parameter estimation with multiple data sets can minimize the offsets between measured and reconciled data. Based on the process data measured from the real plant, numerical results demonstrated effectiveness of application of the proposed method to an air separation process.

Key words: data reconciliation, variable load, steady-state detection, multiple operation conditions clustering

摘要： 过程系统的数据校正与参数估计是进行实时操作优化与过程控制的基础。过程系统变负荷下由于模型参数变化的非线性及显著误差的影响，导致数据校正与参数估计的结果不准确，从而影响实时操作优化与过程控制的效率。针对此问题，本文提出了一种用于变负荷下的数据校正与参数估计方法。此方法主要包括过程的稳态检测与数据采样，多工况下的数据聚类和基于多组测量的数据校正与参数估计。首先选择有效和可靠的过程测量数据，根据变负荷下工况的波动性与系统的非线性特征进行数据聚类，最后基于聚类结果调整模型参数使得模型输出与过程测量数据偏差最小。此方法可有效地减小模型参数变化的非线性及显著误差对数据校正与参数估计结果的影响。基于现场的测量数据，将此方法应用于空气分离流程系统中，结果显示了基于此方法的数据校正与参数估计结果更准确。

关键词: 数据校正, 变负荷, 稳态检测, 多工况聚类