Leveraging Edge Intelligence for Solar Energy Management in Smart Grids

The rapid advancement of renewable energy, particularly solar power, has been significantly supported by the Internet of Things (IoT) infrastructure, enabling optimized energy management in smart grid systems. However, accurate long-term solar energy forecasting remains a critical challenge, especially exploiting specific characteristics of data locality at the edge. This paper introduces an edge intelligence-driven hybrid deep learning model that integrates Temporal Convolutional Networks (TCN) and Gated Recurrent Units (GRU) for precise solar energy prediction. The proposed model leveraged TCN’s capability to capture long-range temporal dependencies and GRU’s efficiency in sequential data processing. To optimize performance for edge computing environments, we employed Neural Architecture Search, model pruning, and quantization techniques, reducing both model size and complexity. The model has been validated using two benchmark datasets (Alibaba Competition and GEFCom), achieving an impressive mean squared error of 0.0051 on long-term forecasts. The results demonstrate that our approach ensures high forecasting accuracy while maintaining computational efficiency, making it well-suited for real-time, IoT-enabled solar energy applications in smart grids.