Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Electrical Load Forecasting over Multihop Smart Metering Networks with Federated Learning

Ratun Rahman, Pablo Moriano, Samee U. Khan, Dinh C. Nguyen

TL;DR

The paper addresses privacy-aware electric load forecasting in multihop smart metering networks by proposing a personalized federated learning framework that mitigates non-IID data across meters via meta-learning to tailor local models. It introduces a personalized meta-LSTM algorithm with optimal per-round learning-rate selection and a latency-optimization scheme for joint resource allocation in leaf and relay nodes, underpinned by a convergence analysis. Empirical results on real-world data show that the proposed PFL approach outperforms traditional LSTM and standard FL in forecasting accuracy while reducing operational latency by up to 45.33%, validating both its predictive quality and responsiveness for smart grid operations. The work advances privacy-preserving forecasting in distributed energy systems and offers a scalable, latency-aware learning paradigm for future smart-grid deployments.

Abstract

Electric load forecasting is essential for power management and stability in smart grids. This is mainly achieved via advanced metering infrastructure, where smart meters (SMs) record household energy data. Traditional machine learning (ML) methods are often employed for load forecasting, but require data sharing, which raises data privacy concerns. Federated learning (FL) can address this issue by running distributed ML models at local SMs without data exchange. However, current FL-based approaches struggle to achieve efficient load forecasting due to imbalanced data distribution across heterogeneous SMs. This paper presents a novel personalized federated learning (PFL) method for high-quality load forecasting in metering networks. A meta-learning-based strategy is developed to address data heterogeneity at local SMs in the collaborative training of local load forecasting models. Moreover, to minimize the load forecasting delays in our PFL model, we study a new latency optimization problem based on optimal resource allocation at SMs. A theoretical convergence analysis is also conducted to provide insights into FL design for federated load forecasting. Extensive simulations from real-world datasets show that our method outperforms existing approaches regarding better load forecasting and reduced operational latency costs.

Electrical Load Forecasting over Multihop Smart Metering Networks with Federated Learning

TL;DR

The paper addresses privacy-aware electric load forecasting in multihop smart metering networks by proposing a personalized federated learning framework that mitigates non-IID data across meters via meta-learning to tailor local models. It introduces a personalized meta-LSTM algorithm with optimal per-round learning-rate selection and a latency-optimization scheme for joint resource allocation in leaf and relay nodes, underpinned by a convergence analysis. Empirical results on real-world data show that the proposed PFL approach outperforms traditional LSTM and standard FL in forecasting accuracy while reducing operational latency by up to 45.33%, validating both its predictive quality and responsiveness for smart grid operations. The work advances privacy-preserving forecasting in distributed energy systems and offers a scalable, latency-aware learning paradigm for future smart-grid deployments.

Abstract

Electric load forecasting is essential for power management and stability in smart grids. This is mainly achieved via advanced metering infrastructure, where smart meters (SMs) record household energy data. Traditional machine learning (ML) methods are often employed for load forecasting, but require data sharing, which raises data privacy concerns. Federated learning (FL) can address this issue by running distributed ML models at local SMs without data exchange. However, current FL-based approaches struggle to achieve efficient load forecasting due to imbalanced data distribution across heterogeneous SMs. This paper presents a novel personalized federated learning (PFL) method for high-quality load forecasting in metering networks. A meta-learning-based strategy is developed to address data heterogeneity at local SMs in the collaborative training of local load forecasting models. Moreover, to minimize the load forecasting delays in our PFL model, we study a new latency optimization problem based on optimal resource allocation at SMs. A theoretical convergence analysis is also conducted to provide insights into FL design for federated load forecasting. Extensive simulations from real-world datasets show that our method outperforms existing approaches regarding better load forecasting and reduced operational latency costs.

Paper Structure

This paper contains 27 sections, 4 theorems, 47 equations, 7 figures, 3 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Our Key Contributions
  3. Paper Organization
  4. Related Works
  5. System Model
  6. Categories of Load Forecasting
  7. Overall System Architecture
  8. Objective Function
  9. LSTM
  10. PFL Algorithm Design for Load Forecasting
  11. Limitations
  12. Convergence Analysis
  13. Latency Analysis For PFL-based Load Forecasting
  14. Formulation of Latency Problem
  15. Proposed Solution to Optimize Latency for the PFL-based Load Forecasting System
  16. ...and 12 more sections

Key Result

Lemma 1

Let Assumption Assump:Variance-gradient hold, the expected upper bound of the variance of the stochastic gradient on local model training is given as $\mathbb{E} ||g_k^j- \bar{g}_k^j||^2 \leq \frac{\sigma_r^2}{N^2}$.

Figures (7)

  • Figure 1: Our proposed architecture for federated load forecasting in the multihop metering network. The SMs network is divided into different routes, each with a subset of SMs in a multi-hop topology. Each SM will train a custom load forecasting model and share the trained model with the utility's server for aggregation.
  • Figure 2: Comparison between different numbers of FL clients (i.e., SMs), standalone, and centralized scheme for IID data.
  • Figure 3: Comparison between different learning rates and meta-learning for 5 SMs non-IID data.
  • Figure 4: Comparison between state-of-the-art approaches (LSTM and FL) and our approach.
  • Figure 5: Simulation result of the original and predicted values for the first 120 minutes in the testing dataset.
  • ...and 2 more figures

Theorems & Definitions (9)

  • Lemma 1
  • proof
  • Lemma 2
  • proof
  • Lemma 3
  • proof
  • Theorem 1
  • proof
  • Remark 1