Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Exploring Lightweight Federated Learning for Distributed Load Forecasting

Abhishek Duttagupta, Jin Zhao, Shanker Shreejith

TL;DR

This paper addresses privacy concerns in load forecasting by replacing centralized learning with Federated Learning on smart-meter data. It introduces a lightweight fully connected neural network trained in a federated setting, augmented with clustering to handle non-IID data and improve generalization. The architecture is a 4-layer feed-forward network with 5569 parameters, implemented using TensorFlow Federated wrappers, and evaluated on the London dataset to achieve an average $RMSE$ of $0.17$ with an IoT energy overhead of $50\ mWh$ on an Arduino platform. Compared with centralized training and prior FL approaches, the method offers privacy-preserving distributed learning with competitive accuracy and reduced computational/communication demands, illustrating its practicality for resource-constrained smart meters.

Abstract

Federated Learning (FL) is a distributed learning scheme that enables deep learning to be applied to sensitive data streams and applications in a privacy-preserving manner. This paper focuses on the use of FL for analyzing smart energy meter data with the aim to achieve comparable accuracy to state-of-the-art methods for load forecasting while ensuring the privacy of individual meter data. We show that with a lightweight fully connected deep neural network, we are able to achieve forecasting accuracy comparable to existing schemes, both at each meter source and at the aggregator, by utilising the FL framework. The use of lightweight models further reduces the energy and resource consumption caused by complex deep-learning models, making this approach ideally suited for deployment across resource-constrained smart meter systems. With our proposed lightweight model, we are able to achieve an overall average load forecasting RMSE of 0.17, with the model having a negligible energy overhead of 50 mWh when performing training and inference on an Arduino Uno platform.

Exploring Lightweight Federated Learning for Distributed Load Forecasting

TL;DR

This paper addresses privacy concerns in load forecasting by replacing centralized learning with Federated Learning on smart-meter data. It introduces a lightweight fully connected neural network trained in a federated setting, augmented with clustering to handle non-IID data and improve generalization. The architecture is a 4-layer feed-forward network with 5569 parameters, implemented using TensorFlow Federated wrappers, and evaluated on the London dataset to achieve an average of with an IoT energy overhead of on an Arduino platform. Compared with centralized training and prior FL approaches, the method offers privacy-preserving distributed learning with competitive accuracy and reduced computational/communication demands, illustrating its practicality for resource-constrained smart meters.

Abstract

Federated Learning (FL) is a distributed learning scheme that enables deep learning to be applied to sensitive data streams and applications in a privacy-preserving manner. This paper focuses on the use of FL for analyzing smart energy meter data with the aim to achieve comparable accuracy to state-of-the-art methods for load forecasting while ensuring the privacy of individual meter data. We show that with a lightweight fully connected deep neural network, we are able to achieve forecasting accuracy comparable to existing schemes, both at each meter source and at the aggregator, by utilising the FL framework. The use of lightweight models further reduces the energy and resource consumption caused by complex deep-learning models, making this approach ideally suited for deployment across resource-constrained smart meter systems. With our proposed lightweight model, we are able to achieve an overall average load forecasting RMSE of 0.17, with the model having a negligible energy overhead of 50 mWh when performing training and inference on an Arduino Uno platform.
Paper Structure (16 sections, 6 equations, 7 figures, 3 tables, 1 algorithm)

This paper contains 16 sections, 6 equations, 7 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Light-weighted FL-based Load Forecasting Scheme
  3. Federated Learning Algorithm
  4. Performance Metrics
  5. DISTRIBUTED LOAD FORECASTING METHODOLOGY
  6. Clustering of smart meter-based energy consumption data
  7. Federated Training and Testing Data sets Creation
  8. Input Window and Output Prediction
  9. Model architecture and TensorFlow Federated Wrappers
  10. CASE STUDY
  11. Case Settings
  12. Results
  13. Performance compared with Centralized Setup
  14. Performance comparison against other research works
  15. Energy consumption of the model on an IoT platform
  16. ...and 1 more sections

Figures (7)

  • Figure 1: Traditional centralized setup for load forecasting.
  • Figure 2: Federated learning setup for load forecasting.
  • Figure 3: Energy profile yearly trend of a random household
  • Figure 4: Comparison of daily consumption over a month: high consumption household vs average consumption household
  • Figure 5: Top to Cluster level Simulation Architecture
  • ...and 2 more figures