A5049139337- Machine Learning and Algorithms
- Fault Detection and Control Systems
- Electric Power System Optimization
- Machine Fault Diagnosis Techniques
- Advanced Optimization Algorithms Research
- Non-Destructive Testing Techniques
- High-Voltage Power Transmission Systems
- Smart Grid Energy Management
- Advancements in Photolithography Techniques
- Advanced optical system design
- Blind Source Separation Techniques
- Energy Load and Power Forecasting
- Optimal Power Flow Distribution
- Advanced Numerical Analysis Techniques
Georgia Institute of Technology
2022-2024
The growing scale of power systems and the increasing uncertainty introduced by renewable energy sources necessitates novel optimization techniques that are significantly faster more accurate than existing methods. AC Optimal Power Flow (AC-OPF) problem, a core component grid optimization, is often approximated using linearized DC (DC-OPF) models for computational tractability, albeit at cost suboptimal inefficient decisions. To address these limitations, we propose deep learning-based...
The transition of the electrical power grid from fossil fuels to renewable sources energy raises fundamental challenges market-clearing algorithms that drive its operations. Indeed, increased stochasticity in load and volatility have led significant increases prediction errors, affecting reliability efficiency existing deterministic optimization models. RAMC project was initiated investigate how move this setting into a risk-aware framework where uncertainty is quantified explicitly...
This paper introduces Dual Interior Point Learning (DIPL) and Supergradient (DSL) to learn dual feasible solutions parametric linear programs with bounded variables, which are pervasive across many industries. DIPL mimics a novel interior point algorithm while DSL classical supergradient ascent. ensure feasibility by predicting variables associated the constraints then exploiting flexibility of duals bound constraints. complement existing primal learning methods providing certificate...
This paper considers optimization proxies for Optimal Power Flow (OPF), i.e., machine-learning models that approximate the input/output relationship of OPF. Recent work has focused on showing such can be high fidelity. However, their training requires significant data, each instance necessitating (offline) solving an To meet requirements market-clearing applications, this proposes Bucketized Active Sampling (BAS), a novel active learning framework aims at best possible OPF proxy within time...