Research On State Estimation And Model Identification Of Dynamic Positioning Ships In Time-varying Environment

Dynamic positioning system plays an important role in the development of modern marine resources and special operations.State estimation and model identification are important components of dynamic positioning system.The accuracy of state estimation and model identification directly affect the overall performance of dynamic positioning system.This paper proposes the corresponding state estimation method and model identification method for the complex time-varying marine environment to improve the accuracy of state estimation and the overall stability of dynamic positioning system.The main work of this paper includes:(1)Aiming at the problem that the statistical characteristics of the process noise of the dynamic positioning ship system in the time-varying environment are unknown and the estimation accuracy of the state estimation method based on the extended Kalman filter decreases,a model predictive extended Kalman filter algorithm is proposed,which estimates the system process noise parameters by comparing the sensor observations and predicted values over a period of time,thereby correcting the system process noise variance in real time and improving the precision of state estimation.In addition,a recursive least squares-based parameter estimation method based on forgetting factor is also studied to estimate the peak frequency of the wave spectrum in a mathematical model of high-frequency motion of a dynamic positioning ship in real time.Simulation results show that when the system process noise is unknown,the model prediction extended Kalman filter can accurately estimate the noise parameters,and the filtering results are more accurate.(2)Aiming at the problem of deviation of state estimation algorithm based on unscented Kalman filtering in the time-varying environment due to abnormal sensor data of the dynamic positioning ship and unknown environmental interference,a robust unscented Kalman filtering algorithm is proposed.The algorithm processes the sensor data anomalies and estimates the unknown environmental forces by adaptively updating the covariance matrix of observation noise and identifying process uncertainty.The simulation results show that when the sensor data is abnormal,the robust unscented Kalman filter can accurately identify the time when the abnormality occurs,and its estimation accuracy and robustness are higher than the conventional unscented Kalman filter.(3)Aiming at the characteristics of non-linear,multi-variable,and strong coupling of the dynamic positioning ship mathematical model,and the problem that conventional genetic algorithms are difficult to control the balance point of convergence speed and identification accuracy when identifying a dynamic positioning ship model,an adaptive recombined genetic algorithm is proposed.This algorithm can effectively improve the quality of identification by introducing a multi-stage sine adaptive genetic operator calculation method and an adaptive recombination replacement strategy,while ensuring the identification convergence speed.Identification simulation comparison experiments verify the effectiveness and superiority of the algorithm.