Skip to main content
EPI Lab

Research Framework

Energy Process Intelligence (EPI) is not a generic AI application for renewable energy. Grounded in Object-Centric Process Mining (OCPM), it focuses on operational processes within and across source–grid–load–storage systems. By linking energy objects, events, constraints, and lifecycles, EPI transforms multi-object operations involving turbines, plants, storage assets, dispatch commands, alarms, work orders, and market transactions into object-centric process representations for discovery, explanation, prediction, and optimization.

Different from adjacent directions

Not generic AI + Energy
The focus is not applying generic models to energy tasks, but building process-aware methods that are explainable, constraint-aware, and verifiable for energy operations.
Beyond traditional power system control
Traditional control emphasizes states, control variables, and stability constraints. Energy Process Intelligence also models business activities, information interactions, and decision behaviors.
Beyond ordinary business process mining
Business process mining mainly studies event sequences. Energy Process Intelligence must also handle physical constraints, multi-object coupling, and safety boundaries.

A coherent chain: representation, explanation, optimization

Representation

Object-Centric Energy Process Mining

Eventize, objectify, process, and constrain heterogeneous energy data.
Explanation

Energy Process Explanation and Risk Analysis

Identify business deviations, physical anomalies, information mismatches, and decision risks through explainable risk analysis.
Optimization

Human-in-the-loop Safe Energy Process Optimization

Generate verifiable and reversible human-in-the-loop decisions under rules, constraints, and safety boundaries.

Research Directions

Direction 01
Object-Centric Energy Process Mining

How can business activities, physical operations, information interactions, and decision-control behaviors involving source-grid-load-storage objects be eventized, object-centric, process-aware, and constraint-aware from heterogeneous data?

Object-Centric Process MiningEnergy Event LogsPhysics-Constrained Process ModelsConformance CheckingEnergy Process Digital Twins
Heterogeneous DataSCADAWork OrdersDispatchingMarketWeatherObject-Centric Process ModelSourceGridLoadStorageEnergy OperationalProcessesPhysics-ConstrainedConformance
Direction 02
Energy Process Explanation and Risk Analysis

How can business deviations, physical anomalies, information mismatches, and decision risks in energy operational processes be jointly discovered, attributed, evaluated, and predictively modeled?

Explainable Process IntelligenceRoot-Cause AnalysisAnomaly EvolutionInformation Mismatch DiagnosisTrustworthy PredictionRule Compliance CheckingRisk Propagation Analysis
Energy Operational ProcessE1E2E3AnomalyRisk Explanation MapBPDRCRootCauseAttribution ExplanationRisk PropagationCompliance CheckTrustworthy Prediction
Direction 03
Human-in-the-loop Safe Energy Process Optimization

How can process knowledge be transformed into verifiable, reversible, and executable human-in-the-loop optimization decisions under operational rules, physical constraints, information trustworthiness, and safety boundaries?

Safe Reinforcement LearningProcess-Aware AgentsHuman-in-the-Loop OptimizationTrustworthy Decision-MakingReversible ExecutionVirtual Power PlantsStorage SchedulingSource-Grid-Load-Storage Operations and Coordination
RollbackProcess CognitionProcess-Aware AgentHuman-in-the-LoopDecisionSafe ExecutionFeedbackSafety Guardrails