A Power Consumption Prediction Method Based on Neural Networks

With the increasing diversity of load types connected to China’s power systems, the power consumption patterns have become increasingly complex. With the increasing complexity of modern power systems, there is a growing demand for more accurate and reliable load forecasting techniques to maintain system stability and security. In response to the deficiencies and insufficient precision of existing prediction methods, this study introduces a gated recurrent unit neural network enhanced through an improved multi-objective particle swarm optimization approach. By integrating swarm intelligence and deep learning, the proposed model achieves enhanced adaptability and improved forecasting performance. The model is validated using seven datasets and compared with five mainstream prediction models. Results show that the proposed model improves prediction accuracy by up to 6.5% over existing methods.