Abstract：Aiming at the slow convergence speed and easy to fall into the local extremum of traditional neural network, an improved wavelet neural network is proposed in this paper to realize the global optimization of network. Firstly, a novel wavelet functional link neural network (NW-FLNN) is constructed by combining wavelet neural network with random vector functional link net. Secondly, the wavelet basis function is introduced into the hidden layer of neural network, and the gradient correction method is used to train the network model. Finally,the New South Wales state electricity price data of Australia is selected as the experimental dataset, and the prediction performance of the new wavelet neural network, the back propagation neural network and ordinary wavelet neural network are compared, respectively. The experimental results show that the new network prediction model has better performance than BP neural network and wavelet neural network. It can significantly reduce the number of network iteration and the number of hidden neurons, and the average percentage error is reduced by 0.0317, which meets the real-time requirements.