Robust adaptive backstepping active control of compressor surge based on wavelet neural network

Abstract

Surge active control can extend the stable operating range of the compressor. Most of the previous active control algorithms can only work on the basis that compression characteristics are known, which limits the engineering application of it. The robust control can alleviate the negative effects brought by uncertain system characteristics, however its designed control law is too conservative that active control can't give full play to its benefits. To overcome the weakness of the former active controllers, this paper comes up with the idea of active surge control, which assumes that the characteristics of the compression system are unknown. And it uses the wavelet neural network algorithm, a kind of algorithm with the nature of approaching any linear and nonlinear functions, to approach the unknown compressor steady-state characteristics asymptotically. Then, by applying backstepping on the second-order MG compressor surge dynamic model with CCV, a robust adaptive active controller of compressor surge is designed, which makes the control law effective to a larger extent, and greatly improve its performance in tracking and anti-interference. From the simulation results, it can be seen that the designed controller displays the strong robustness against un-modeled dynamics, flow and pressure disturbances, and in the meanwhile achieves near to zero asymptotical tracking error. The method proposed in this paper allows compressors to operate stably with high-pressure ratio and high-efficiency beyond the surge boundary, namely greatly expands its operating range. Thus, this paper has significant guidance for future development, theoretical research, and engineering application of surge active control of the compressor.