A modified subspace identification method and its application in thermal process

作者：Wang Xin^*; Lü Jianhong; Xiang Wenguo

来源：Proceedings of the CSEE, 2010, 30(32): 96-102.

• Error-state
• Numerical algorithms
• State-space
• Subspace identification
• Thermal process

A modified subspace identification method (SIM) was proposed according to the drawbacks of MOESP (multiple-input multiple-output output-error state space model identification) and N4SID (numerical algorithm for subspace state space system identification) in capturing system matrices (A, B, C, and D) of a state-space model. Taking the advantage that the matrices A and C could be directly calculated from the observation matrix, the proposed method calculated the matrices A and C earlier using MOESP, and then the matrices B and D were calculated by N4SID later. Accordingly, the method not only need not to construct large matrices needed in MOESP for calculating B and D, but also can reduce the computational complexity, due to avoidance of solving the linear least square problem needed in N4SID for calculating A and C. Then, the method was applied to the identification of a gas fired power plant and Alstom gasifier model. Some performance criteria such as CPU runtime, floating-point operations (FLOPS) and model relative error of the method were introduced to the comparison between the modified SIM, MOESP and N4SID. The simulation results show that the modified SIM could both promise the identification precision and improve the calculation efficiency. Especially under large data condition, the FLOPS and CPU runtime could be effectively reduced.

• 出版日期2010
• 单位东南大学