Embedded Files
Research Projects and Plans
Smart grid situational awareness under uncertainty,
Risk-aware decision-making in power systems,
Multi-fidelity and multi-resolution sensor data analytics,
Cybersecurity in energy systems,
Heterogeneous data analyses from various components of smart cities.
Research Experiences
Post-Doctoral Research
Post-Doctoral Research
Smart-Meter Data Analytics:
Analyzing the smart meter load consumption data to understand the regular, seasonal, and abnormal energy consumption behavior and their dependencies on various factors. Applications: Energy theft detection, false data injection detection, and localization using smart meter data.
Generalized Horizon PV Forecasting in Multi-Site Scenario:
GSP and GNN techniques for analyzing spatio-temporal weather data and renewable energy (solar and wind) data for power output prediction and their impact on electric grids.
Multi-Fidelity GNN for Power Flow Analysis
GNN for Smart Infrastructures
Ph. D. Research
Ph. D. Research
Detection, location identification of cyber and physical stresses in smart grids:
Different cyber and physical stresses are modeled considering the time-series representation of power grid measurements.
Extracting features from the instantaneous state correlation matrix of the power grid for detecting and locating stresses.
Graph signal processing techniques are being used for better modeling of topologically distributed power system data and enhancing detecting and locating efficiency. In particular, two novel GSP-based techniques: local smoothness second time-derivative (LSSTD) method and vertex-frequency energy distribution (VFED)-based method have been developed for the detection and location identification of stresses.
Fig.: Local Smoothness of bus voltage graph signal in the IEEE 118 bus system : (a) normal condition, (b) false data injection attack at bus 100.
Characterization and classification of stresses:
Contribution 1: A two-stage classification framework for cyber and physical stresses in smart grids:
Stage 1: Binary classifier to classify between the cyber and physical stresses.
Stage 2: Classification among different types of cyber and physical stresses depending on the predicted binary class in the first stage.
Classification between physical stresses: abrupt load change vs. transmission line failure.
Classification among cyber stresses: DoS, FDIA, replay attack, ramp attack, delay attack.
Contribution 2: Classification and Characterization of Cyber Attacks:
Binary classification: Multiple random attacks Vs. Clustered attacks
For Clustered cyber attacks, estimating attack center and attack radius.
Combination of GSP-based feature extraction and machine learning-based techniques for characterization, and classification of stresses.
Fig: Different types of cyber attacks on voltage angle time series.
Recovery of states in smart grids:
Missing state recovery techniques for smart grids have been developed using graph signal sampling-reconstruction framework as well as using the correlation among the states.
A novel state recovery technique based on the global and local smoothness of power system graph signal has been proposed which is agnostic of the bandwidth of the graph signal.
Other Works in Ph.D. :
A GSP-based analysis of the effect of a single bus perturbation in the electric grid.
Optimum placement of measurement devices in smart grids has been studied under graph signal sampling-reconstruction framework.
Surveying on interaction graphs for cascading failure analysis in power grids.
Designing edge-computing framework for smart grids.
Page updated
Google Sites
Report abuse