The invention discloses a wind speed prediction method based on a complex numerical value forward neural network, and the method comprises the steps: obtaining a training set and a prediction set for wind speed prediction, constructing the complex numerical value forward neural network, and initializing a parameter vector; introducing a GroupLasso regularization item into a target function trained by the complex numerical value forward neural network, converting training into solution of a constraint optimization problem, training the complex numerical value forward neural network by adopting a training set and a customized complex numerical value projection quasi-Newton algorithm, and ending training until a preset condition is met to obtain a trained complex numerical value forward neural network, and inputting the prediction set into the trained complex numerical value forward neural network to obtain a wind speed prediction result. According to the method, the GroupLasso regularization item is introduced, and the complex numerical value forward neural network is trained by using the customized complex numerical value projection quasi-Newton algorithm to realize optimization of the network structure and parameters, so that the network structure is compact, the generalization performance is high, and meanwhile, the accuracy of wind speed prediction is improved.