Research Statement

It is undeniable that electric power is one of the most powerful and critical technologies that led to rapid industrialization and globalization in the twentieth century. The electric power grid is over a century old and is considered the largest and most complex interconnected physical system. The ever-increasing integration of intelligent IoT devices has forced researchers and practitioners to re-evaluate power grids' reliability, safety, and security. This research addresses power system state estimation, one of the critical components of wide-area situational awareness (WASA). WASA comprises automatic prevention, perception, and restorative actions of any anomalies directed towards the smart grid. Thus, to ensure proper WASA, it is essential to provide the status information of the grid to the control center, such that anomalies can be detected and addressed immediately. Smart grids rely on the deep integration of cyber components, which provides invaluable opportunities for a more secure and reliable intelligent grid operation. However, the critical interdependency of power grids on the cyber elements, introduces new vulnerabilities toward cyber and physical attacks. The immense volume of energy data collected by various sensors, such as Phasor Measurement Units (PMUs), provides new opportunities for detecting, estimating, and predicting multiple events in the system using data analytics techniques and machine learning. Understanding the nature of attacks and exploiting the vulnerabilities is very important to designing a proper defense against the smart grid's ever-increasing cyber and physical stresses. Some key objectives of the research includes:

  • Optimally partition the grid to implement multi-area distributed and data-driven estate estimation algorithms that will require no or reduced topological information (admittance matrix, physical description, and locations of the components) to monitor and predict the status of each system state effectively.

  • Promptly detect and localize the component failure or cyber stresses by predicting steps ahead of system states using machine learning techniques.

  • Using data-driven system discovery, develop state estimation algorithms to recover masked or unobservable states of the system to retain full system observability.

Using the distributed mode of operation will reduce the overhead communication burden, thus improving overall latency for time-critical functions such as detection and localization of anomalies. Discovering complete system observability from partial measurements will allow optimal and economically feasible deployment of expensive computational and monitoring hardware such as PMUs, edge computing devices, etc.

Research Projects

Grid Integration of Renewable Energy & Microgrid Testbed Design

To meet the demand of the next generation power system, renewable energy resources can be the fuel of choice because it is readily available, free of cost, environment-friendly, and the renewable energy-based generation is cost-effective in all manners. Several types of renewable energy resources, such as solar, wind, geothermal, tides, and biomass. This project's concentration is limited to solar energy resources, solar plants, wind turbines, natural gas generators, and storage systems to provide required power support. In particular, this project is associated with the mathematical modeling of the renewable sources and simulation of the different aspects and cases for various scenarios. Besides that, the Zinc Bromide Battery and Li-ion Battery are delineated with explanations of their performance and related simulations. After that, both for the cases of islanded mode and grid-tied mode operation, the microgrid systems and the storage unit's performance are analyzed for different parameters. All the results are verified by MATLAB simulations meticulously.


Related Publications:

[1] Jakir Hossain, N. Sakib, E. Hossain, R. Bayindir, "Modelling and Simulation of Solar Plant and Storage System: A Step to Microgrid Technology", International Journal of Renewable Energy Research (IJRER), Volume 7, Issue 2, Pages 723-737.

[2] E. Hossain, Jakir Hossain, N. Sakib, R. Bayindir, "Modelling and Simulation of Permanent Magnet Synchronous Generator Wind Turbine: A Step to Microgrid Technology", International Journal of Renewable Energy Research (IJRER), Volume 7, Issue 1, Pages 443-450.

[3] N. Sakib, Jakir Hossain, E. Hossain, R. Bayindir, "Modelling and Simulation of Natural Gas Generator and EV Charging Station: A Step to Microgrid Technology", International Journal of Renewable Energy Research (IJRER), Volume 7, Issue 1, Pages 399-410.

[4] J. Hossain, S. S. Sikander, and E. Hossain, "A wave-to-wire model of ocean wave energy conversion system using MATLAB/Simulink platform," 2016 4th International Conference on the Development in the in Renewable Energy Technology (ICDRET), Dhaka, 2016, pp. 1-6.

[5] Hossain, Eklas, Feng Shi, Ramazan Bayindir, and Jakir Hossain. "Feasibility Analysis: Evaluating Sites for Possible Renewable Energy Options and Their Implications to Minimize the Cost." The International Journal of Electrical Engineering & Education, (June 2020).

[6] N. Sakib, J. Hossain, H. I. Bulbul, E. Hossain, and R. Bayindir, "Implementation of unit commitment algorithm: A comprehensive droop control technique to retain microgrid stability," 2016 IEEE International Conference on Renewable Energy Research and Applications (ICRERA), Birmingham, 2016, pp. 1074-1079.

[7] I. Colak, E. Hossain, R. Bayindir and J. Hossain, "Design a grid-tie inverter for PMSG wind turbine using FPGA & DSP builder," 2016 IEEE International Power Electronics and Motion Control Conference (PEMC), Varna, 2016, pp. 372-377.




Wide-area Situational Awareness: State Estimation, Detection & Localization of Cyber-physical Intrusions in Smart Grids

Wide area situational awareness (WASA) comprises automatic prevention, perception, and restorative actions of any anomalies directed towards the smart grid. Thus, to ensure proper WASA, it is customary to make available the status information of the grid to the control center such that detected anomalies are immediately addressed. Smart grid operations rely heavily on the seamless integration among cyber and physical elements. Due to this critical interdependency on the cyber aspects, modern power grids exhibit new vulnerabilities to cyber and physical attacks, for example, intentional manipulation of sensor measurements and masking the physical failure. However, the immense volume of energy data collected by various sensors, such as Phasor Measurement Units (PMUs), also provides new opportunities for detecting, estimating, and predicting multiple events in the system using data analytics techniques and machine learning. Thus, understanding the nature of attacks and exploiting the vulnerabilities is very important to designing proper defense against the smart grid's ever-increasing cyber and physical stresses. This project aims to develop privacy-aware distributed state estimation algorithms that can promptly detect and localize cyber-physical intrusions utilizing the PMU data. Thus, improving the overall situational awareness of the smart grid.


Related Publications:

[1] M. J. Hossain and M. Rahnamy-Naeini, "State Estimation in Smart Grids Using Temporal Graph Convolution Networks," North American Power Symposium (NAPS 2021), 2021, pp. 01-05, Texas, USA.

[2] M. J. Hossain and M. Rahnamy-Naeini, "Data-Driven, Multi-Region Distributed State Estimation for Smart Grids," IEEE PES Innovative Smart Grid Technologies, Europe 2021 (ISGT Europe), 2021, pp. 1-6 Espoo, Finland.

[3] M. J. Hossain and M. Rahnamy-Naeini, "Line Failure Detection from PMU Data after a Joint Cyber-Physical Attack," 2019 IEEE Power & Energy Society General Meeting (PESGM), Atlanta, GA, USA, 2019, pp. 1-5.

[4] M. A. Hasnat, M. J. Hossain, A. Adeniran, M. Rahnamay-Naeini, and H. Khamfroush, "Situational Awareness Using Edge-Computing Enabled Internet of Things for Smart Grids," 2019 IEEE Globecom Workshops (GC Wkshps), Waikoloa, HI, USA, 2019, pp. 1-6.

Privacy-aware Advanced Metering Infrastructure: Dynamic Pricing, Demand Response, and Non-Intrusive Load Monitoring

In this project, non-intrusive load monitoring (NILM) function in AMI is considered to demonstrate the promising potential of EI and FL in supporting smart grid applications. NILM allows detailed energy sensing from aggregated data (e.g. from smart meters), which is essential for energy management solutions and addressing energy conservation challenge due to increasing energy demands. However, NILM can reveal detailed appliance-specific energy consumption statistics, which can expose privacy and security risks to consumers and, as such, needs to be handled delicately. In this paper, 2-tier and 3-tier FL frameworks are developed for the NILM function. The performance of the proposed models are evaluated with respect to communication cost and the model training loss. We compare these approaches to a centralized deep learning solution. Our results show that our 2-tier and 3-tier FL approaches show comparable performance to the centralized deep learning approach. However, our 2-tier and 3-tier FL approaches reduce the associated communication cost, while providing some immediate layer of privacy, because raw data is not communicated to central servers in AMI systems. The presented study along with recent developments in this domain suggest that EI and FL have great potential in addressing data processing challenges in smart grids. However, research and practice on this emerging multidisciplinary domain—EI-enable smart grids—are still in a very early stage. These studies can help bridge the gap between highly dynamic and fast-growing EI and FL domains and smart grids. The marriage of these technologies can ultimately work towards a more reliable, secure, and efficient smart grids.


Related Publications:

[1] N. Hudson, M. J. Hossain, M. Hosseinzadeh, H. Khamfroush, M. Rahnamay-Naeini, and N. Ghani, "A Framework for Edge Intelligent Smart Distribution Grids via Federated Learning," 2021 International Conference on Computer Communications and Networks (ICCCN), 2021, pp. 1-9.

Graph-Based Machine Learning in Smart Grid: Grid Partitioning, Graph Neural Network, Cascading Failure Analysis

The dynamics of power systems are governed by various physical and operational attributes of these systems. The underlying interactions and interconnections among the components of power systems are important attributes that their effects are reflected in the structured data collected from these systems. The structural topology of the power system and the embedded structures in the data due to interactions among the components are valuable information that can help with data-driven SE methods. However, many of the existing data-driven SE techniques do not adequately take into account such information. To fill this gap, in this paper, temporal as well as graph-based features of power system measurements are considered in analyzing PMU time-series for SE. Specifically, a Graph Convolutional Neural Network (G-CNN) is combined with GRU units to create a spatio-temporal model that uses the topology of the system in learning the patterns embedded in the PMU time-series for SE. The performance of the presented approach is compared with the SE techniques based on Minimum Mean Square Error (MMSE), Bayesian Multivariate Linear Regression with Auto Regression, Support Vector Regression, and Multivariate LSTM, which do not explicitly consider the graph structures in data. The presented SE approach is also evaluated for the cases when the measurements are available from all the buses and also when the measurements are only available from a subset of buses. The presented SE method shows improved performance under both cases compared to the SE techniques that do not explicitly consider the graph structures.


Related Publications:

[1] M. J. Hossain and M. Rahnamy-Naeini, "State Estimation in Smart Grids Using Temporal Graph Convolution Networks," North American Power Symposium (NAPS 2021), 2021, pp. 01-05, Texas, USA.

[2] M. J. Hossain and M. Rahnamy-Naeini, "Data-Driven, Multi-Region Distributed State Estimation for Smart Grids," IEEE PES Innovative Smart Grid Technologies, Europe 2021 (ISGT Europe), 2021, pp. 1-6 Espoo, Finland.

[3] Nakarmi, Upama; Rahnamay Naeini, Mahshid; Hossain, Md J.; Hasnat, Md A. "Interaction Graphs for Cascading Failure Analysis in Power Grids: A Survey" Energies 13, 2020.

Watts Going On (WGO) : IoT Device Smart Non-Intrusive Load Monitoring

A student project supervised by me: The goal of the WGO team is to develop an IoT device that can be plugged into the electric meter. The device can efficiently identify and monitor the energy consumption of individual household appliances without being connected to individual appliances via a process called "Non-Intrusive Load Monitoring". The disaggregated energy data then will be communicated to an interactive support application directly to the consumer's smartphone.