Apparatus and method for automated data selection in model identification and adaptation in multivariable process control

摘要

A computer-based apparatus and method for automated data screening and selection in model identification and model adaptation in multivariable process control is disclosed. Data sample status information, PID control loop associations and internally built MISO (Multi-input, Single-output) predictive models are employed to automatically screen individual time-series of data, and based on various criteria bad data is automatically identified and marked for removal. The resulting plant step test/operational data is also repaired by interpolated replacement values substituted for certain removed bad data that satisfy some conditions. Computer implemented data point interconnection and adjustment techniques are provided to guarantee smooth/continuous replacement values.

主权项

1. A method of screening and selecting data automatically for model identification and model adaptation in a multivariable predictive controller (MPC), the method comprising:
given an online controller having an existing model, in a processor: loading process data from a subject process and storing said process data in a database accessible by said model; using a rule-based data selector, automatically detecting and excluding data segments of the stored process data that are unsuitable for model quality estimation and for model identification, the detecting and excluding comprising the steps of:
(a) collecting process data variables and storing collected process data variables in said database at a given sampling frequency as time series variables,(b) loading data status, special values, and value limits of variables of the subject process with their corresponding time series from the database,(c) screening a given time series variable as a dependent process variable or an independent process variable and applying basic data screening filters to detect and mark data segments of the time series as Good Data or Bad Data according to given data quality measurement parameters,(d) grouping time series variables according to their associated proportional-integral-derivative (PID) loops,(e) comparing process variables (PV) against their corresponding set points (SP) in each PID loop and applying data screening filters,(f) generating predictions for dependent variables without a PID association using available independent variable measurements,(g) evaluating said generated predictions for dependent variables without a PID association against corresponding available dependent variable measurements and applying data screening methods, and(h) identifying and generating Bad Data slices in the given time series variable using the data segments marked as Bad Data and a data slice generator to exclude said Bad Data segments from the time series; validating the excluded data segments to minimize the data loss from bad data segments being excluded; repairing certain data segments of said stored process data to maximize usage of data in a MPC application; and updating said existing model using resulting process data as stored in the database.