An innovative application for energy consumption forecasting in smart grid integrating BiLSTM with DBN

With the continuous increase in the global energy demand, enhancing energy management and utilization efficiency has become crucial. This study proposes a novel energy prediction model integrating the Gaussian restricted Boltzmann machine and bidirectional long short-term memory (BiLSTM) networks. This model algorithm is dedicated to predicting continuous energy data, using deep belief networks for feature extraction and deep feature representation learning, while efficiently capturing time series correlations of energy data through BiLSTM. In the experiment analyzing smart meter data of UK residents, the algorithm demonstrated excellent performance in long-term and short-term energy consumption prediction, significantly outperforming traditional models such as recurrent neural network, convolutional neural network, and long short-term memory. Comparative experimental analysis has proven that this algorithm has significant advantages in processing large-scale energy data and mastering time series characteristics, reflecting its important role in practical prediction work.