Process Model Identification Using Historical Data for MPC
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for identifying process models for model-based predictive multivariable control in process plants require significant effort and expertise, particularly in planning and implementing active tests to generate sufficient measurement data.
Innovation Solution
A computer-implemented method for automated identification of process models using previously defined controlled, manipulated, and disturbance variables, which involves providing historical measurement data, determining operating points, and systematically exciting manipulated variables to generate informative measurement data, followed by model identification using the least squares method.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If step tests are used to excite dynamic processes for model identification, then measurement data with sufficient information content can be generated, but the planning and implementation effort increases significantly and requires deep expertise in control engineering
Solution Approach 1:
The system performs self-identification by automatically analyzing historical measurement data to generate process models without requiring external step tests or expert intervention. The automated algorithm selects appropriate data periods, performs model fitting, and validates results independently, eliminating the need for manual test planning and implementation while maintaining model quality.
Solution Approach 2:
The system uses pre-existing historical measurement data from normal plant operation to perform model identification before any active testing is required. By analyzing routinely collected data with appropriate filtering and preprocessing, the system prepares sufficient information for accurate model identification in advance, avoiding the need for subsequent step tests.
2Reliability
If intensive discussions between plant operator and control engineering service provider are held to plan step tests, then adequate test plans can be developed, but the time and resource investment increases
Solution Approach 1:
The automated identification system independently performs all planning and execution of the identification process without requiring human discussions or manual test plan development. The system automatically selects historical data periods, determines appropriate model structures, and executes the identification algorithm, eliminating time-consuming collaborative planning sessions while maintaining or improving plan quality through systematic automated decision-making.
3Measurement precision
If manual model identification is performed with expert knowledge, then accurate process models can be obtained, but the effort and expertise requirements are high
Solution Approach 1:
The system performs automated model identification using algorithms that independently analyze historical measurement data, select appropriate model structures, and optimize parameters without requiring expert knowledge or manual intervention. This automation maintains model accuracy through systematic mathematical optimization while dramatically simplifying the operator's task to merely initiating and monitoring the automated process.
Solution Approach 2:
The system replaces manual expert analysis and judgment with automated computational algorithms. Instead of relying on human experts to interpret data and select model parameters, mathematical optimization algorithms and automated selection criteria perform these functions, substituting mechanical human effort with computational processing while maintaining or improving consistency and accuracy.
Data Source
Figure 1
Figure 2
AI summary
A computer-implemented method is proposed for the automated identification of a process model for model-based, predictive multivariable control of a process plant. This method utilizes previously defined controlled variables, manipulated variables, and disturbance variables for the model-based, predictive multivariable control of the process plant. The method comprises: - providing historical measurement data from a production operation of the process plant in an archive, where the manipulated variables were constant during production operation; - determining the respective operating point of all manipulated variables, as well as the respective operating point and standard deviation of all controlled variables from the historical measurement data; - specifying a permissible deviation of each controlled variable from the operating point of the respective controlled variable, where the permissible deviation is, in particular, six times the standard deviation of the respective controlled variable; - sampling the controlled variables.Manipulated variables and disturbance variables with a constant sampling time, - Provision of a respective low-pass filter for the controlled variables, whereby a filter time constant of the low-pass filter is chosen such that the standard deviation of the respective controlled variable with the low-pass filter is smaller by a factor of 2 to 6, preferably 3 to 5, than without the low-pass filter, - Further steps are carried out stepwise for each manipulated variable.