A Federated Learning Algorithm That Combines DCScaffold and Differential Privacy for Load Prediction

Accurate residential load forecasting plays a crucial role in optimizing demand-side resource integration and fulfilling the needs of demand-side response initiatives. To tackle challenges, such as data heterogeneity, constrained communication resources, and data security in smart grid load prediction, this study introduces a novel differential privacy federated learning algorithm. Leveraging the federated learning framework, the approach incorporates weather and temporal factors as key variables influencing load patterns, thereby creating a privacy-preserving load forecasting solution. The model is built upon the Long Short-Term Memory (LSTM) network architecture. Experimental results demonstrate that the proposed algorithm enables federated training without the need for sharing raw load data, facilitating load scheduling and energy management operations in smart grids while safeguarding user privacy. Furthermore, it exhibits superior prediction accuracy and communication efficiency compared to existing federated learning methods.