Abstract

This study applies the generative adversarial network-based optimization approach to site selection and capacity determination of energy storage device in a power grid. Through the combination of modified long short-term memory and generative adversarial networks, the proposed method enhances the learning capability for the decision support of energy storage allocation. This method excels at the utilization of modified long short-term memory to ensure a better data-generation and data-discrimination in a generative adversarial network, enabling the achievement of effective data learning and deduction. To validate the feasibility of the proposed approach, a practical system as well as an example system are both examined under different scenarios, where the placement cost, peak load, and voltage deviation are all concerned. Test results gained from this study are beneficial for energy storage industry applications. In this study, a novel approach is proposed for site selection and capacity determination of energy storage devices in power grids by applying a generative adversarial network-based optimization method. The proposed approach combines modified long short-term memory and generative adversarial networks to enhance the learning capability for decision support of energy storage allocation. Specifically, the modified long short-term memory improves the data-generation and data-discrimination in the generative adversarial network, leading to effective data learning and deduction. To demonstrate the feasibility of our proposed approach, a practical system as well as an example system are tested under different scenarios, where the placement cost, peak load, and voltage deviation are all taken into considerations. Test results indicate the feasibility of the method, providing valuable insights for the energy storage industry.