Regional Power Demand Forecasting Using BiLSTM and ConvBiLSTM Models for Sustainable Energy Management

The purpose of this study is to develop an electricity demand forecasting model to improve the efficiency of energy management in specific regions. To achieve this, this study integrates various renewable energy sources and large-scale energy storage systems (ESS) to ensure the stability of the power grid. In this study, regional electricity demand forecasting is conducted using climate data (air temperature and ground temperature) for two major cities in Jeju Special Self-Governing Province of South Korea, Jeju-si and Seogwipo-si. Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), and Convolutional BiLSTM (ConvBiLSTM) are used as forecasting models to compare their performance, and the forecasting performance of each model is evaluated to more accurately reflect the variability of electricity demand due to climate change. The performance results show that in Jeju-si, which has a relatively stable demand pattern, the model combining BiLSTM and ConvBiLSTM performs the best, enabling it to effectively learn the time dependencies and complex patterns. However, in Seogwipo-si, where demand volatility is comparatively high, the models performed relatively poorly, suggesting that more data and complex modelling approaches are needed. Consequently, this study highlights the potential of using deep learning models in renewable energy-based local power management, particularly for reducing power losses and increasing grid reliability through locally produced energy consumption.