Continuous health-based optimal operation and control for mission-critical power systems

• Main Author: Kumar, Logesh
• Other Authors: Hung Dinh Nguyen
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering; Turn-to-turn short circuit; Fault managment; Finite element method; Winding function approach
• Online Access: https://hdl.handle.net/10356/180090
• Institution: Nanyang Technological University

The mission-critical systems such as marine and aerospace comprises of prime mover, rotating machine, energy storage system, and power switching devices to deliver the required load power. Due to the adverse operating condition and external stress, the electrical power system is prone to various faults in its components. Timely detection of these faults can mean the difference between continuous operation and unforeseen system failure. Therefore, detection and post fault operation are of paramount importance to accomplish the set mission, where redundant power system may not always be available. In recent years’ lot of research has been conducted to detect the faults in the system, but very few and little work is reported on the mitigation and post fault scenario. All these formulations handle equipment failure contingencies in binary fashion i.e., either perfectly working or completely failed. In this thesis, the formulation of health is defined as a single continuous function. Further, the health values are used in system operation to dispatch optimal load power, and in control strategy to reduce the effect of fault in the component. The analytical machine modelling with faults is carried out using high-fidelity winding function approach method. The finite element method-based machine modelling with faults is carried out in ANSYS. The frameworks developed in this thesis are verified using the simulation models. An experimental test-rig is built with provision to introduce turn-to-turn fault to validate the proposed frameworks. A health condition index is theoretically defined based on physics of fault in the first part of the study for various faults in permanent magnet synchronous machine, and battery system. In addition, the relationship between the health condition index and the variation of real power and voltage due to fault characteristics is obtained based on simulation data. The upper bound of the asset is updated in real-time based on fault characteristics to dispatch the optimal current. However, modifying the bounds based on fault characteristics result in reduced power output. Therefore, in the second part of the thesis an optimal fault management framework is proposed for continuous operation until the mission is accomplished, and maintenance triggered to reduce the stress on the asset. The turn-to-turn short circuit (TTSC) current injection method lead to the initialization of irreversible demagnetization, and further degrade the winding insulation. Therefore, a new framework is proposed for optimal current injection during TTSC fault condition with objective of maintaining the remaining useful life (RUL) until the generator maintenance is triggered. RUL calculation based on aging model is proposed in this thesis, which considers the aging factors due to electrical, mechanical, and thermal phenomena. A bottom-up approach through the developed model is used to estimate the lifetime of the machine in real-time. In addition to the TTSC fault effect on magnetic flux linkage, the demagnetization fault effect overlap resulting in reduced estimation accuracy. Therefore, the data-driven approach, and model-based approach, which depends on the fault data to train the machine-learning algorithm, and for estimation of fault parameters deteriorate. Therefore, a novel framework is proposed in this thesis to compensate the demagnetization characteristics, which includes real-time update of the magnet’s knee point, and a demagnetization fault indicator that reflects the initialization of the irreversible demagnetization state. The machine-learning algorithm result confirm that data integrity is improved, and time-domain signal analysis are sufficient for training and diagnosis of TTSC fault severity. A model-based approach, based on moving horizon estimation (MHE) with control reference, and current and voltage residuals are used. The proposed estimation method can accurately estimate the TTSC fault parameters for different load conditions under constant frequency generator operation and can identify a low severity TTSC fault of 5% for any low severity of insulation resistance. A novel way to include the thermal constraint is presented in this thesis considering the magnetic density limitation on the knee point of magnet, along with proposed demagnetization fault indicator and magnet temperature limitation. A safe operating region as presented in first part of the thesis is utilized in the generator capability curve, and variation due to electrical and thermal limits are updated in real-time for three-stage successive relaxation optimization to dispatch optimal current. The result confirms the availability of generating unit with 36% increase in remaining useful life for 5% short with fault severity of 1 Ω. Further, the proposed framework can operate the machine at rated and lower power rating based on machine condition without compromising on the load.