Energy Efficiency Prediction of Energy Storage Virtual Synchronous Machine Based on Long Short-Term Memory Network

Abstract

The integration of renewable energy sources in the power grid has led to an increased demand for energy storage systems to manage the intermittency and variability of these sources. Among various energy storage technologies, Energy Storage Virtual Synchronous Machines (ESVSMs) have emerged as a promising solution for enhancing grid stability and energy efficiency. However, optimizing the performance of ESVSMs requires accurate energy efficiency predictions. This paper proposes a energy efficiency prediction network (EEPNet) to achieve energy efficiency prediction of ESVSMs. Firstly, considering the peculiarity of storing virtual synchronous machine efficiency fluctuations, this paper improves the traditional Inception structure. This allows for the extraction of information at different scales and its fusion to obtain a better representation. Secondly, this paper introduces residual connections in the structure. This not only avoids the difficulties in model training caused by deep networks but also helps to integrate the original information with the extracted information. Thirdly, building upon the improved Inception structure, this work combines long short-term memory (LSTM) and attention mechanisms to construct a complete network architecture, further enhancing the performance of the entire model. Finally, the effectiveness and validity of the model are validated through comprehensive experiments.