Implementation of performance assessment tool for multivariable, model-predictive controller

Abstract

The aim of the thesis was to implement a tool for performance assessment of a multivariable, model-predictive controller which was in this work NAPCON Controller developed by Neste Jacobs. The aim of control performance assessment and monitoring is to ensure that control systems operate as required. In practice, control performance assessment techniques are usually based on a comparison of the current controller performance and a benchmark value defined by some criteria. The result of this comparison is called the control performance index. In this thesis, the technological performance of the controller was measured with two techniques using different criteria for calculating the benchmark value. These selected methods were historical and design-case benchmarks. In addition, the economic performance of the controller was assessed.

In this work, an OPC UA database was used for storing the calculated performance indices as well as the related configuration parameters. This required the definition of new OPC UA based information models. The implemented performance assessment tool included a performance calculation application as a Windows service and a graphical user interface for configuring the performance assessment calculations.

The functionality of the implemented controller performance assessment tool was tested with a simulator of a distillation unit and against an actual MPC controller. The results of the different simulation cases showed that the calculated performance indices responded as expected when the process conditions or the control objectives changed. The tool requires some additional testing and development before it can be deployed to a real process environment as a part of the controller software, although the created performance assessment tool worked well according to the simulation results.