Federated Learning for Load Forecasting Improved by New Initialization Strategies

This paper addresses challenges in federated short-term load forecasting (STLF) caused by heterogeneity in client load data. The researchers propose two model initialization strategies: a pretrained initialization using auxiliary public data to reduce client drift, and a sequential local initialization (SLIAvg) that adapts models progressively within communication rounds. These methods are compatible with existing federated learning frameworks and have demonstrated improved forecasting performance, convergence, and reduced errors in experiments with real smart-meter data. AI

IMPACT Introduces new initialization techniques to improve the performance and convergence of federated learning models in critical infrastructure applications like load forecasting.

RANK_REASON Academic paper detailing a novel methodology for a specific machine learning task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• federated learning
• Short term load forecasting with recency effect
• SLIAvg

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →