Effective demand response and GANs for optimal constraint unit commitment in solar-tidal based microgrids

Abstract

A new approach for optimal demand response program (DRP) in the microgrid considering the high penetration of the solar energy and tidal units as significant and popular renewable sources in the system is proposed here. The proposed method makes use of a multi-objective problem (MOP) to not only minimize the total operation cost of the scheduling problem but also mitigate the high risk of the interruption in power delivery due to the components failure rate and long repairing rates. Considering the high complexity and nonlinearity of the formulation, a novel heuristic method based on the firefly algorithm is introduced to solve the problem without any assumption or killing the accuracy. In addition, a dynamic three-phase correction (DPC) formulation is proposed which can help to increase the global search characteristics of the method when boosting the convergence capability of the model. Due to the hard predictability nature of the solar irradiance, a deep learning model based on generative adversarial networks (GAN) is presented to predict the output power of the solar and tidal units properly. The high performance and feasibility of the proposed multi-layer problem are assessed on an IEEE test system.