Neural network ship overall model approximation method based on improved gradient descent method

The invention discloses a neural network ship overall model approximation method based on an improved gradient descent method, and the method comprises the steps: taking the storm interference as a part of a ship, and taking the storm interference as an overall model; performing approximation on the overall model through a neural network; and adopting an improved gradient descent method to carry out online updating on the weight, and performing simulation on the ship according to an approximated ship model or enabling the method to be used for designing a ship motion controller. According to the method, external interference of the ship and the ship are regarded as a time-varying overall model, through neural network approximation, an improved gradient descent method is adopted to update the weight on line, and a better approximation effect is achieved. According to the method, the change of the ship model caused by the change of loading capacity, draught, external interference and the like in the sailing pro