How can cyber-physical energy systems operate autonomously while preserving safety and reliability?
My research combines physical modeling, data-driven learning, optimization and control, high-fidelity simulation, and operational data to develop methods for reliable autonomous energy-system operation.
RESEARCH ARCHITECTUREclosed loop
z={zAMI,zSCADA,zweather,zDER}
Measurements and operational evidence
p(x,G∣z)
State, topology, behavior, and uncertainty
x^k+1=fθ(x^k,uk,dk)
Physical structure with adaptable learned dynamics
u⋆=argminuJ(x,u)s.t.g(x,u)≤0
Constrained decisions, coordination, and control
Pgrid+PPV+PBESS=Pbuilding+PEV
Grid, buildings, PV, batteries, EVs, and physical constraints
3Φ · BALANCED
CURRENT RESEARCH
Current Research
AREA 01
Building Dynamics & Scientific Machine Learning
I develop control-oriented building thermal models spanning RC-network and structured grey-box models, data-driven models, parameter estimation, and scientific machine learning. Current questions include transferability across operating conditions and whether improvements in model accuracy translate into better control and grid-edge performance.
Thermal dynamicsEstimationSciMLTransferability
CzdtdTz=RoaTout−Tz+Qint+QHVAC
AREA 02
Community Energy Management & Intelligent Control
My work investigates MPC and reinforcement learning for residential and community energy systems with HVAC, rooftop PV, battery storage, and flexible loads. SmartCommunitySim provides a configurable environment for evaluating energy-management controllers in both grid-connected and outage operation.
MPCRLHVACPV + Battery
AREA 03
Distribution-Grid Modeling & Intelligence
My current work also includes distribution-system simulation and analytics using topology, asset, and operational data, with emphasis on model inference, topology validation, AMI analytics, and uncertainty-aware understanding of distribution systems.
TopologyAMIModel inferenceUncertainty
AREA 04
Simulation & Co-Simulation
I use building simulation, residential-community simulation, and building–grid co-simulation to evaluate models, optimization methods, and control strategies in closed loop. These environments bring together weather, grid-edge resources, electric-network behavior, control algorithms, and communication effects within a unified experimental framework.
My future research program consists of four independent but connected thrusts directed toward reliable autonomous grid operation.
Auto exploring research thrusts
RELIABLEAUTONOMOUSGRID OPERATION
THRUST 01
Transferable AI & Scientific Machine Learning for Building Energy Flexibility
Can we learn building models once and adapt them reliably across populations of buildings rather than repeatedly learning each building from scratch?
Transfer learningCross-building adaptationCross-climate generalizationControl-oriented learningPhysics-guided learning
THRUST 02
Safe & Scalable Multi-Agent Coordination of Heterogeneous DERs
How can heterogeneous energy resources coordinate at scale while preserving local constraints, reliability, and autonomy?
MPCDistributed optimizationMulti-agent RLSafety filtersHybrid model-based + learned control
THRUST 03
Multi-Layer Building–Grid Co-Simulation for Closed-Loop Validation
How should autonomous energy-system algorithms be evaluated when buildings, DERs, the electric network, controllers, and communication systems interact in closed loop?
Building dynamicsDERsDistribution gridControllersCommunication / cyber layer
THRUST 04
Uncertainty-Aware Grid Intelligence from Operational Data
How can operational data reveal hidden grid structure and behavior while explicitly representing what remains uncertain?
AMI + smart-meter dataSCADA / feeder measurementsWeather + DER dataAsset + topology recordsProbabilistic inferenceUncertainty quantification
3Φ · BALANCED
VALIDATION PHILOSOPHY
Validation Philosophy
I do not consider an algorithm trustworthy because it performs well on a single dataset or simulation.
Auto exploring validation evidence
Real Operational Data
Tests robustness to measurement noise, missing information, heterogeneity, and actual system behavior.
High-Fidelity Simulation
Enables controlled experiments where field testing is impractical or unsafe.
Closed-Loop Controller Testing
Tests whether models remain useful once decisions alter the physical system.
Cross-System Generalization
Tests whether methods capture transferable structure rather than memorize one operating environment.
3Φ · BALANCED
LONG-TERM VISION
Long-Term Vision
My goal is not to eliminate human operators or engineering judgment. I seek to develop cyber-physical energy systems capable of making increasingly sophisticated decisions with limited supervision while preserving safety and reliability.
Reliable autonomy will require learning, physical models, optimization, control, simulation, and operational data to work together—not a single algorithm in isolation.
