Historical data driven identification for multivariable systems based on state observation and teaching-learning-based optimization algorithm
DONG Ze
YIN Er-xin
Abstract：The conventional multivariable system identification method based on the combination of intelligent algo-rithms and historical data selects historical data, which represent the system from steady-state to dynamic-state, as model-ing data. When the modeling data contain unknown disturbance, this method cannot establish the correct system model. Therefore, a historical data driven identification method for multivariable systems based on state observation and teaching-learning-based optimization algorithm is proposed. In this method, historical data representing the system changing from dynamic-state to steady-state are treated as modeling data. The steady-state component is removed based on final steady-state value. Then the data are divided into two segments. The system status at the end of the first segment is obtained by means of state observer and prediction model, then it serves as the initial system status of the second segment. Input data of the second segment and the prediction model are employed to simulate the system. And in order to make the simula-tion output close to the actual output, teaching-learning-based optimization algorithm is adopted to optimize the prediction model parameters. In the modeling simulation of a multivariable system, the result shows that the method can overcome the disturbance effect on the precision of model identification. Finally, the coordinated control system modeling of a thermal power unit is carried out, and simulation results show the method effectiveness.
Keywords：disturbancestate observerteaching-learning-based optimization algorithmmultivariable systemshistor-ical data drivenidentification
Publication Date：2017-01-01
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：11( 1369-1379 )