Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Federated Learning Forecasting for Strengthening Grid Reliability and Enabling Markets for Resilience

Lucas Pereira, Vineet Jagadeesan Nair, Bruno Dias, Hugo Morais, Anuradha Annaswamy

TL;DR

The paper addresses cyber-physical threats in DER-rich grids by combining federated learning (FL)–based probabilistic forecasting with a hierarchical local electricity market (LEM) to detect and mitigate attacks. FL forecasts enable per-prosumer anomaly detection by comparing forecast errors to feeder injections, enabling threshold-based DoS-like attack identification, while the LEM clears dispatch using an ACOPF framework that incorporates load flexibility and resilience scores. Upon detection, the market coefficients are updated using $ Delta = P_{PCC}-ar{P}_{PCC}$ to steer dispatch toward local DER and flexible loads, demonstrated by updating $(oldsymbol{\alpha}_i, oldsymbol{eta}_i, oldsymbol{\xi})$ to $(oldsymbol{ ext{primed}})$ and redispatching. In a Madeira LV test case with 88 nodes, a PV shutdown attack caused feeder imports to rise, but the mitigation reduced imports to about 29 kW, roughly a 40% improvement over the unmitigated scenario, illustrating practical gains in grid reliability and resilience. The work also highlights future directions for scalable FL training, heterogeneous data handling, and incorporating forecast uncertainty into robust or stochastic market formulations.

Abstract

We propose a comprehensive approach to increase the reliability and resilience of future power grids rich in distributed energy resources. Our distributed scheme combines federated learning-based attack detection with a local electricity market-based attack mitigation method. We validate the scheme by applying it to a real-world distribution grid rich in solar PV. Simulation results demonstrate that the approach is feasible and can successfully mitigate the grid impacts of cyber-physical attacks.

Federated Learning Forecasting for Strengthening Grid Reliability and Enabling Markets for Resilience

TL;DR

The paper addresses cyber-physical threats in DER-rich grids by combining federated learning (FL)–based probabilistic forecasting with a hierarchical local electricity market (LEM) to detect and mitigate attacks. FL forecasts enable per-prosumer anomaly detection by comparing forecast errors to feeder injections, enabling threshold-based DoS-like attack identification, while the LEM clears dispatch using an ACOPF framework that incorporates load flexibility and resilience scores. Upon detection, the market coefficients are updated using to steer dispatch toward local DER and flexible loads, demonstrated by updating to and redispatching. In a Madeira LV test case with 88 nodes, a PV shutdown attack caused feeder imports to rise, but the mitigation reduced imports to about 29 kW, roughly a 40% improvement over the unmitigated scenario, illustrating practical gains in grid reliability and resilience. The work also highlights future directions for scalable FL training, heterogeneous data handling, and incorporating forecast uncertainty into robust or stochastic market formulations.

Abstract

We propose a comprehensive approach to increase the reliability and resilience of future power grids rich in distributed energy resources. Our distributed scheme combines federated learning-based attack detection with a local electricity market-based attack mitigation method. We validate the scheme by applying it to a real-world distribution grid rich in solar PV. Simulation results demonstrate that the approach is feasible and can successfully mitigate the grid impacts of cyber-physical attacks.
Paper Structure (12 sections, 1 equation, 5 figures)

This paper contains 12 sections, 1 equation, 5 figures.

Table of Contents

  1. Introduction and background
  2. Federated Learning
  3. Methods
  4. Using FL Forecasts to Identify Cyber Attacks
  5. Local Electricity Markets
  6. Mitigation Approach Using Load Flexibility
  7. Simulation Case Study
  8. Low Voltage Distribution Grid
  9. Consumption and PV Production Data
  10. Federated Forecasting
  11. Experiments and Results
  12. Conclusions and Future Work

Figures (5)

  • Figure 1: Overview of hierarchical LEM, with the PM and SM layers utilized for resilience.
  • Figure 2: LEM co-located with distribution grid. This shows a primary and secondary feeder distribution network based on the modified IEEE-123 node test case.
  • Figure 3: PV generation before the attack.
  • Figure 4: Net total feeder power injection (3-phase).
  • Figure 5: Distribution of load curtailment across nodes.