Closed-loop Identification Of Thermal System Based On Subspace

The modern thermal power unit is developing toward the unit plant.The unit plant itself poses more stringent requirements for the design of the controller because of the large inertia,strong coupling,large difference in the dynamic characteristics between the boiler part and the steam turbine part.Large-scale integration of new energy has higher requirements for the thermal power units to achieve rapid variable load capacity.Traditional control algorithms,such as PID,can't meet the requirements of automatic control of large thermal power units.Advanced control strategies often rely on precise system models.Therefore,system identification is particularly important.System identification is evolving with automatic control.Traditional identification algorithms,such as the classical least-squares algorithm and its variants,are difficult to apply well in large thermal power units,mainly due to the characteristics of thermal power units.Therefore,it is very important to propose a new system identification algorithm that can get a good identification result on systems with large inertia and large delay.The subspace identification algorithm can get the state space equation of the system directly from the input and output data,and is widely used in the identification of multivariable systems because of its good numerical stability and low computational complexity.Begin with the basic theory of subspace identification algorithm,I deduce the process of classical subspace identification algorithm including the classical algorithm MOESP,CVA into a unified framework,and give a detailed simulation example.Then,I summarize the existing subspace identification algorithms systematically,and for the first time classify them into principal component analysis class and projection operation class,and gives a detailed theoretical proof process.During the process of combing existing subspace algorithms,the author proposed a new instrumental variable algorithm.And make a systematic simulation analysis of these algorithms under the same data.Then the author combines the classical subspace identification algorithm with the traditional two-stage identification algorithm,and gives the comparison between the fused algorithm and the traditional two-stage identification algorithm.In the end,this paper uses the operation data in and step experiment data of the thermal power plant,and combines the classical open-loop subspace identification algorithm and the subspace identification algorithm of the instrumental variables proposed in this paper to make a detailed analysis,and analyzes the effect of data preprocessing on the experimental results.Overall,the subspace identification algorithm can more accurately estimate the dynamic characteristics of the system,which has its reference value in the selection of practical engineering algorithms.