Hybrid deep learning techniques for providing incentive price in electricity market

This paper proposes a novel deep learning approach for the incentive price prediction in the electricity market. The proposed model makes use of the wavelet transform to first decompose the signal and then the decomposed signals are employed for training a deep learning model based on generative adversarial networks (GANs). Such a procedure would reduce the complexity of the data and help the GAN model to train more efficiently and optimally. In order to adjust the GAN model parameters, a novel evolutionary algorithm based on the adaptive particle swarm optimization (APSO) is developed. The proposed APSO uses an adaptive structure to update the weighting factor and the social acceleration coefficients in a way that the convergence rate of the algorithm would increase and the global search ability of the algorithm boosts. Two different case studies are used to examine the appropriate performance and feasibility of the proposed hybrid model. The simulation results advocate the efficiency and accuracy of the proposed hybrid method. © 2022