Thesis - Open Access
Master of Science (MS)
Department / School
Electrical Engineering and Computer Science
Timothy M. Hansen
Inverter-based generation is replacing synchronous-based generation in the power grid. In the past, due to the dominance of synchronous-based generation, inverter-based generation dynamics were typically under-modeled as they made little contribution to voltage and frequency control. Looking at the present context, several disturbances, like the Blue Cut Fire (CA), on August 16, 2016, the South Australian Blackout on September 28, 2016, and the Canyon Two Fire Disturbance (CA) on October 9, 2017, etc., are caused by the influence of converter-based resources. So, for converter-dominated power system stability analysis and control, the dynamics of an inverter-based system cannot be under-modeled. In addition, grid operators, researchers, system planners, and other stakeholders require realistic IBRs dynamic models to understand how power systems will operate at distinct layers of IBRs penetration, what criteria should be used to determine grid code requirements, and what compliance requirements should be included. Furthermore, control applications may require model approximations when the user has no information about the system. For instance, online techniques that estimate model states and parameters require complete information about the system. The primary objective of this thesis is to assess the model approximations of an unknown system, considering it as a black box model, which balances model accuracy with computation cost. This thesis prioritizes the dynamic modeling of commercially available inverters (assuming it as a black box) due to their growing connection to the modern power grid. It employs an active probing signal-based data-driven modeling technique to extract the dynamics model of a smart photovoltaic (PV) inverter operating in Volt-VAr, Volt-Watt, and Freq-Watt modes following the IEEE 1547-2018 standard. In addition, the impact of varying irradiance on dynamics models of a photovoltaic inverter (PV) will also be analyzed along with a detailed explanation of the design criteria of perturbation signal required during dynamic modeling. The proposed approach was successfully verified with an experimental setup using OPAL-RT (real-time digital simulator) and power-hardware-in-the-loop (PHIL) . Additionally, inverter-based systems have low inertia and are more vulnerable to significant rate-of-change-of-frequency (ROCOF) and frequency deviations. The growing use of renewable energy sources also leads to uncertainties in time-varying system parameters such as inertia and damping constants, which are more challenging in microgrids dominated by converters and low inertia. Hence, several controllers are employed to limit the ROCOF and frequency within an acceptable limit. Accurate information about the system states and parameters is essential for adequately functioning control mechanisms, but measuring all states using sensors may not be practically feasible. Furthermore, system parameters change over time due to aging, varying operating points, etc., and sensor data is often noisy, leading to degraded control and monitoring processes. Therefore, states and parameters must be estimated to monitor the grid properly. Several estimators, like the Kalman filter and its family (extended Kalman filter, unscented Kalman filter), moving horizon estimation, particle filters, etc., can be used as estimation techniques. However, choosing filters that balance accuracy and computation time is essential, particularly for fast frequency control mechanisms. Moreover, during estimation, the system can be perturbed using an excitation signal, which is generated from an energy storage system (ESS) and whose suitability depends on the topology of the ESS. In this thesis, the second objective involves applying the designed perturbation signal from the first part to assess the characteristics, computation time, and accuracy of various estimation filters. The study will evaluate and compare different estimation algorithms using a frequency dynamics system. This study mainly focuses on estimating three power system states (frequency, power angle, and rate-of-change-of-frequency) and two power system parameters (inertia constant and damping constant) using various estimation filters, and a characteristic, computation time requirement, and accuracy level comparison is made among them. The main outcomes of this thesis are to extract a data-driven dynamic model for commercial off-the-shelf (COTS) inverters and analysis of varying irradiance on those obtained models. These models can be employed by researchers, system planners, and stakeholders to perform simulation analyses. The aim is to mitigate disturbances arising from incorporating distributed energy resources (DERs). Furthermore, the thesis emphasizes implementing estimation filters in a real-time environment using Opal-RT real-time simulator, allowing users to adapt these filters based on computational time or precision requirements.
South Dakota State University
Poudel, Bidur, "Dynamic Modeling of Commercial-Off-The-Shelf Smart Inverters Using Power-Hardware-In-The-Loop" (2023). Electronic Theses and Dissertations. 701.
Available for download on Thursday, August 15, 2024