Investigating the performance of RNN model to forecast the electricity power consumption in Guangzhou China

Abstract

The project initiatives to create a reliable prediction model for power loads in Guangzhou, China. The power industry is facing issues due to rapid market growth and the necessity for better grid management, prompting this response. In developing the models, conventional machine learning models have been used so far, but in this study, the performance of deep learning is investigated. Therefore, the recurrent neural network (RNN) was selected for the prediction of electricity consumption. Later, the performance of the model was compared with autoregressive integrated moving average (ARIMA), long short-term memory (LSTM), and RNN. The experimental results show that the RNN outperforms ARIMA and LSTM, with an R² value of 0.92, an RMSE of 0.13107 and an MAE of 0.0176. The project improved power resource planning and management, selected an acceptable forecasting model RNN and contributed to forecasting technology developments. The study identified limits in historical data availability and quality, as well as external influences affecting the studies. RNN models can help optimize resource allocation and improve energy planning.