Short-term Wind Power Forecasting Based on Conditional Generative Adversarial Networks

Abstract

Accurate short-term wind power prediction can effectively assess the fluctuation characteristics of wind power in the short term in the future, which is important for formulating reasonable scheduling plans, collaboratively solving grid frequency regulation problems and ensuring safe and stable grid operation. Due to the randomness and volatility of wind power timing, existing forecasting methods fail to fully learn the deep relationships implied in nonlinear power data, resulting in lower prediction results. A short-term wind power prediction model based on conditional generative adversarial network is proposed in this paper. The proposed model uses a convolutional neural network (CNN) to form the internal structure of the generator and discriminator, takes the wind power influencing factors as conditions, and introduces the characteristic loss function as the loss function of the hidden layer of the discriminator. Then, the generator is trained to generate predicted power data conditional on the power influencing factors through the game of conditional generative adversarial network, so as to achieve short-term wind power prediction. This paper uses data from a centralised wind power station in north-west China to demonstrate the algorithm and compare the prediction results of the proposed model with those of other models to verify that the proposed method can improve the prediction accuracy of short-term wind power while taking into account the generalization ability.