Industrial

March 2020 - Present -- Smart Vehicle-Grid Integration

• The project aims to study the impacts of Level-3 EV charging on a utility distribution network and develop a strategy to mitigate those impacts

March 2020 - September 2020 --- High Performance Computing-based Dynamically Adaptive Protection Schemes for Electric Grid

• An improvement over the state-of-the-art protection scheme over static pre-determined or device health based protection settings to a less conservative protection and remedial action schemes to mitigate cascading failures.

March 2020 - Present -- Irrigation Modernization

• The Irrigation Modernization project is about bringing enhancements to agriculture water distribution systems. INL is building an analytics engine that computes the economics of adding hydropower, ecosystem services, converting canals to pipes, adding PV arrays, etc.

June 2019 - August 2019 -- Customized model generation of smart buildings for efficient grid enhancement services using Machine Learning and data driven approaches

• Validate physics-based model with real field data to evaluate the percentage error which can be better resolved by developing a customized building specific model using Variational Auto Encoder (VAE) and Generalized Adversarial Networks (GAN). The methodology involved dealing with large dataset for multiple buildings which represented a general category. Data augmentation and learning methods were utilized to develop a predictor engine model which could generate power consumption of the buildings without the availability of training dataset. Methods like LSTM were utilized in the development of time series predictor model, but the performance feature of VAE outperformed the other underlying models.

May 2018 - August 2018 -- DARPA SD2 (Deep Learning Algorithm Development)

• Development of a new algorithm using Deep Learning skills associated with different distance preservation techniques when projecting a data from a high dimension to a lower dimensional space, using global distance preservation between data points. Biological experiments have many variables considering their sample, laboratory environment etc. These experiments are costly to be performed on a repetitive basis, the goal was development of a baseline algorithm which through deep learning, reinforcement learning, could predict such outcomes of large dataset of experiments and compare it with one such performed experiment.

June 2017 - August 2017 -- Stalling of Induction Motors in Fault-Induced Delayed Voltage Recovery (FIDVR)

• Development of a PNNL taxonomy feeder with composite load models for commercial buildings and designed a protection scheme in FORTRAN to reduce the effect of FIDVR conditions due to the stalling of Induction Motors. The test system was designed and tested in PSCAD.

December 2013 - July 2015 -- Automatic credit card statement generation invoking HPExstream engine

• Given the responsibility for design and development of automated credit card statement generation. The process involved data cleansing process utilizing some ETL tools like Informatica and data preparation through SQL.