(364s) Dynamic Modeling and Estimation for Condition Monitoring of Energy Systems

Research Interests: process monitoring, advanced process control, optimization, hybrid modeling

Software Skills: MATLAB, JMP, Aspen Hysys, Python, Aspen Custom Modeler

My research develops dynamic modeling and estimation techniques for operating plants to improve process monitoring and control. The integration of intermittent renewables into the power grid has necessitated frequent load adjustments, causing damage, reduced reliability, and increased maintenance costs for critical equipment’s. To address these issues, my work aims to create advanced condition monitoring tools that can help plants understand load-following effects. The main objectives are to equip utility plants with tools for efficient preventive maintenance, outage prevention, and advanced control strategies, enhancing operational flexibility while ensuring safety and reliability. During my PhD research, I actively collaborated with my fellow graduate students and leading research agencies like the Electric Power Research Institute (EPRI) and corporate partners such as Southern Company. I organized and participated in plant visits, presented findings in weekly meetings, prepared comprehensive reports, and delivered impactful presentations, enhancing our research outcomes while developing strong professional relationships and communication skills.

• Dynamic Modeling

My work focuses on the development of advanced and adaptive condition monitoring tools for industrial processes using comprehensive dynamic process models that accurately represent system geometry and real-time behavior across various operational scenarios. I specialize in developing detailed first principles models based on ordinary and partial differential equations (ODEs and PDEs) to capture the dynamic behavior of energy systems. These models have been rigorously validated using industrial datasets from diverse energy systems, including coal-fired and combined cycle plants. Sensitivity studies using two years of operating data have shown that the models align well with real-time behavior and effectively capture system non-linearities. This expertise forms a strong basis for creating advanced and adaptive condition monitoring tools customized for industrial processes.

• State Estimation

In my research, I investigate the development of advanced monitoring techniques for joint state and parameter estimation in energy systems using dynamic models. To address the challenges posed by incomplete and inaccurate data obtained from operating plants, I have developed optimal quantitative estimation algorithms. These algorithms include applying dynamic models alongside Kalman filter-based estimation techniques for adaptive, real-time condition monitoring of critical high temperature components found in various energy systems.

Furthermore, to address the limitations of existing estimation algorithms available in literature while using detailed dynamic models. I have developed modified estimation approaches to handle uncertainties in both differential and algebraic equations, addressing noise in both states effectively. As part of my current research work, I am also developing advanced constrained state estimation algorithms to ensure adherence to mass and energy balances in process systems, crucial for handling sensor data that may violate fundamental principles. Validating these algorithms with literature and industrial data ensures accuracy and robust integration when using sensor data with process models in a comprehensive model-estimation frameworks. To summarize, dynamic modeling and estimation techniques offer promising advancements in energy system monitoring, utilizing abundant data for enhanced accuracy and comprehensiveness

• Grey- Box Modeling

Process systems are typically modeled using two approaches: white-box models, using dynamic first principles models, and black-box models, using operational measurement data. While white-box models offer predictive power, they can struggle with complex, physics-unknown phenomena like flow maldistribution and unsteady heat transfer commonly encountered in energy systems. They are also computationally intensive and less adaptable for real time monitoring applications. In contrast, black-box models based on data-driven techniques are easier to build and adapt online but may lack predictive accuracy, especially with extrapolation or incomplete measurement data.

My research aims to develop modeling approaches that integrate first-principles and data-driven methods to monitor complex dynamic process systems effectively. This includes developing innovative grey-box models combining rigorous physics-based models with data-driven approaches, particularly valuable for systems with unknown phenomena or poorly understood physics. I specialize in developing distributed differential-algebraic equation (DAE) models to analyze high-temperature boiler components like superheaters and reheaters under load-following conditions. These models ensure accuracy in heat transfer calculations and system characterization through mass and energy balances. A modular design approach has been adopted in these models, making them versatile for various systems. Additionally, I specialize in developing data-driven black-box models using machine learning techniques that adapt online using operational data.

As a future outlook, I am using hybrid grey-box modeling approaches to create a robust health monitoring framework for energy systems. By combining extensive plant data with physics-based models, we can improve monitoring accuracy and reliability. Additionally, I am working to innovate grey-box architectures that integrate serial, parallel, or integrated structures for modeling multi-scale process systems. These developments have potential applications in equipment life consumption analysis and unconventional process control.

Publications :

• Saini, V., Bhattacharyya, D., Purdy, D., Parker, J., and Boohaker, C. (2024). Nonlinear state estimation of a power plant superheater by using the extended Kalman filter for differential algebraic equation systems. Applied Thermal Engineering, 251, 123471. https://doi.org/10.1016/j.applthermaleng.2024.123471
• Beahr, D., Saini, V., Bhattacharyya, D., Seachman, S., & Boohaker, C. (Accepted for publication). Estimation-based model predictive control with objective prioritization for mutually exclusive objectives: Application to a power plant. Journal of Process Control.
• Mukherjee, A. *, Saini, V. *, Adeyemo, S.*, Bhattacharyya, D., Purdy, D., Parker, J., & Boohaker, C. (Under review). Development of hybrid first principles-artificial intelligence dynamic models for applications to power plant monitoring under load-following operation. Energy Conversion and Management (*Equal Contribution).
• Saini, V., Bhattacharyya, D., Purdy, D., Parker, J., & Boohaker, C. (Internal review). Development of a hybrid grey-box modeling framework for health monitoring of high temperature boiler components. To be submitted to Applied Energy.