基于神经网络的北太平洋柔鱼渔场预报

Forecasting on fishing ground of red flying squid (Ommastrephes bartramii) in the North Pacific Ocean based on artificial neural net

根据1998-2004年6-11月份我国鱿钓生产数据(月份、作业船数、经纬度和日产量)以及对应的海洋环境因子数据,即5 m水层的海水温度、46 m水层的海水温度、112 m水层的海水温度、317 m水层的海水温度、叶绿素a含量以及海平面高度距平值等,以经标准化后的单位捕捞努力量渔获量(CPUE)作为中心渔场指标,采用多种BP神经网络预报模型,对北太平洋柔鱼渔场进行了分析与比较。通过对13种神经网络预报模型的比较,以及实际CPUE的验证,以拟合残差最小的预报模型作为最优预报模型,认为结构为9-7-1的BP神经网络模型相对误差仅为0.008 570,可作为北太平洋柔鱼渔场的预报模型。

Based on the fishing data including months, the number of vessels, the catch per day of the red flying squid （Ommastrephes bartramii ） in jigging fishery by Chinese fishing fleet, and the data of the corresponding oceanic environment, i.e. longitude, latitude, the water temperature at 5 m （ T005 ） , 46 m （T046）, 112 m （Tl12）, 317 m （T317） under surface, chlorophyll a （ CHA）, and sea surface height anomaly（SSHA） in the North Pacific Ocean during June and November in 1998 to 2004, the BP neural networks model was applied to predict the emergence and distribution of fishing grounds of red flying squid after the standardization of CPUE. The total of 13 models with different hidden layers have been tested. The optimum model was selected by comparing the values of the simulation residual of each model structure statistically. The result shows that the BP neural networks model with 9-7-1 networks structure with 0. 008 570 simulation residual only can be used for better predicting the fishing ground of the red flying squid in North Pacific Ocean.