Privacy-preserving collaboration
Learn across renewable-energy sites without pooling raw measurements. Separate data locality from additional protections for exchanged parameters.
FedDRL · Wind power forecastingAn academic research agenda
Trustworthy Federated AI for Resilient Energy Systems
How can energy systems learn together when data must remain local and participants cannot be trusted by default?
FedTrust.ai brings together research on collaborative learning, privacy protection and robustness for cyber-physical energy systems.
Explore the researchResearch vision
Collaboration with
accountable trust.
Energy data are distributed across organisations, assets and operating conditions. Sharing model updates can enable collaboration, but keeping raw data local is only a starting point.
Our research connects task-specific learning with explicit privacy mechanisms and checks on unreliable updates. The longer-term goal is to understand how these methods can support dependable sensing, forecasting and decisions under disruption.
01 / Research themes
From local data to shared intelligence.
Four connected lines of enquiry, grounded in published work.
Robust learning with unreliable data
Combine missing-data reconstruction with verifiable privacy mechanisms and trust-aware aggregation to address anomalous client updates.
ZTFed-MAS2S · Wind data imputationDistributed cyber-threat detection
Train collaborative detectors for false data injection attacks, pairing Transformer models with Paillier encryption in federated learning.
Secure federated learning · FDIA detectionFederated energy decision-making
Share learning across microgrids using multiagent reinforcement learning and physics-informed rewards for cost and energy self-sufficiency.
F-MADRL · Multimicrogrid management02 / Selected publications
Evidence behind the agenda.
Four journal articles connecting federated methods to energy-system tasks.
2026Published
Data imputation · Verifiable privacy · Trust-aware aggregation
ZTFed-MAS2S: A Zero-Trust Federated Learning Framework With Verifiable Privacy and Trust-Aware Aggregation for Wind Power Data Imputation
IEEE Transactions on Industrial Informatics, 22(1), 165–175.
Combines attention-based sequence-to-sequence imputation with verifiable differential privacy, parameter-transmission checks and dynamic trust-aware aggregation, evaluated on wind-farm datasets.
2024Published
Federated reinforcement learning · Energy management
Federated Multiagent Deep Reinforcement Learning Approach via Physics-Informed Reward for Multimicrogrid Energy Management
IEEE Transactions on Neural Networks and Learning Systems, 35(5), 5902–5914.
Shares agent parameters across microgrids while retaining local operational data, using physics-informed rewards to address operating cost and energy self-sufficiency.
2023Published
Collaborative forecasting · Local data retention
Wind power forecasting considering data privacy protection: A federated deep reinforcement learning approach
Applied Energy, 329, 120291.
Integrates DDPG-based forecasting with federated parameter sharing for ultra-short-term wind prediction, enabling collaboration without centralising raw wind-farm data.
2022Published
False data injection · Encrypted federated learning
Detection of False Data Injection Attacks in Smart Grid: A Secure Federated Deep Learning Approach
IEEE Transactions on Smart Grid, 13(6), 4862–4872.
Couples Transformer-based detectors with federated learning and the Paillier cryptosystem, evaluating collaborative attack detection on IEEE 14-bus and 118-bus test systems.
Publication years refer to the journal volume. Preprints are linked separately from the published versions.
03 / Research framework
Connect the method to the task.
A conceptual map of the research, with system-level resilience as a continuing goal.
Published methodEvaluated taskResilience question to pursue
Transformer + encrypted FLPaillier-protected collaboration
False data injection detectionIEEE test systems
Can reliable detection improve timely operational response?
ZTFed-MAS2SPrivacy verification + trust-aware aggregation
Missing wind-data reconstructionWind-farm datasets
Can robust reconstruction sustain downstream decisions during data loss?
Evidence boundary. These studies provide task-level experimental evidence. They do not establish an integrated operational platform or field-deployed grid resilience.
Future direction. Evaluate how privacy, robustness and model utility interact with physical constraints and resilience outcomes across sensing, prediction and decision-making.
Different mechanisms, different guarantees.
Local data retention limits raw-data sharing; encryption protects specified computations or exchanges; differential privacy bounds disclosure under a defined mechanism; robust aggregation addresses unreliable updates. Their assumptions and guarantees must be assessed separately.
04 / Academic contact
Let’s connect on
trustworthy energy intelligence.
Prof. Yang Li
School of Electrical Engineering
Northeast Electric Power University
Jilin, China