Research And Application Of Target Drift Prediction Model For Maritime Search And Rescue

With the implementation of "Maritime Power" and "the Belt and Road" strategies,the frequencies of the active oceanic operations as well as the risk of marine casualty are considerable in China,so the issue of maritime safety has attracted increasing attention.In order to timely respond to increasingly frequent maritime accidents and protect the safety of people’s lives and property,relevant departments around the world have attached great importance to maritime search and rescue work.Target drift prediction is the premise of Marine search and rescue(SAR)operation,and it is also the most complex and critical part of SAR work.In order to avoid the blindness of the organization of SAR operations and improve the efficiency and accuracy of the command and coordination of maritime SAR,it is very important to study the prediction model of maritime target drift.The drift prediction of maritime search and rescue target mainly involves two key technologies: the drift calculation model of maritime target and the search area calculation model of maritime target.The drift calculation model of maritime target mainly simulates its trajectory by analyzing the response law of target drift motion to marine atmospheric environment dynamics,while the search area calculation model of maritime target calculates the search area by analyzing the trajectory simulation error on the basis of trajectory simulation.In order to study the above two models respectively,the tracking and observation data of unpowered drift of a typical offshore fishing vessel in the Pearl River Estuary area of the South China Sea and the drifting buoy data of the West Coast of the United States were combined to conduct the experiments in this paper.The specific work is as follows:1.Based on the unpowered drift experiment of a typical offshore fishing vessel in the Pearl River Estuary of the South China Sea,two models,the AP98 model and the semi-analytical model of drift dynamics,were studied in this paper.For the semianalytical model of drift dynamics,the model coefficients were determined by using multiple linear fitting rates according to the experimental data.Then,the least square method was used to obtain the nine wind-induced drift coefficients in the AP98 model.The Jibing phenomenon of the target fishing boat was studied,and the Jibing frequency were obtained.In addition,the divergence angle of wind-induced drift and the probability of positive crosswind(POPC)of offshore fishing boats during the experiment period were also calculated,and based on this,an improvement of AP98 model was proposed.Finally,Lagrange particle tracking and Monte Carlo techniques combined with several different model parameter schemes are used to simulate the drift trajectory and search area of the target fishing vessel.The advantages of the AP98 model over the semi-analytical model of drift dynamics were verified by experimental comparison,and the effective improvement of the AP98 model was verified by considering the POPC coefficient.2.Based on the data of drifting buoys in the west coast of the United States,the search area calculation model of maritime target was studied in this paper.The environmental dynamic field data(the wind field was from ERA5 data set and flow field was from HYCOM)combined with drifting buoy track data were used to fit the optimal wind-induced drift coefficient of drifting buoy.Then,the experimental sea area was divided into 69 custom grids.Two sub-grid velocity models,the random flight model and the random walk model,were established respectively.The velocity standard deviation,the time scale and the diffusion coefficient of Diffusion and Advection Simulation Errors(DASE)in each grid were calculated respectively.In addition,based on the two components of DASE velocity,a class of sub-grid velocity model based on ARMA model was proposed in this paper.Unlike the two classical sub-grid velocity models,the ARMA model predicts the DASE velocity with a certain step length by analyzing the relationship between the current value and the historical value of the time series,which can further study the time correlation of the sub-grid velocity series.Finally,a set of comparison experiments combined with the kernel density estimation method were used to calculate the search prediction region under the 95% confidence interval of the simulated particle set.Through the evaluation and comparison of the experimental results,the performance of the random walk model and the ARMA model,as well as the effective influence of the regional variation of diffusion coefficient on the accuracy of the random walk and random flight model were verified.3.Taking the HF ground wave radar data in the Pearl River Estuary of the South China Sea as the experimental background and taking the drift of unpowered fishing vessels with typical characteristics as the experimental object,the sub-grid velocity models and the improved AP98 model were further studied.In the experiment,the DASE velocity was calculated by fitting the surface velocity observed by the ground wave radar and the velocity observed by the ADCP.The experimental results show that considering the POPC coefficient in the search area calculation can reduce the average prediction distance error to a certain extent,thus improving the prediction accuracy.Compared with the random walk model,the simulation results of the random flight model can better cover the real target trajectory,but the average prediction error of the random flight model is larger.Compared with the two classical sub-grid velocity models,the ARMA model has a certain improvement in the average prediction accuracy of the search area and the probability of successfully covering the search target trajectory in the simulated prediction search area.