Search

This paper clearly demonstrates the structure of Experimental Advanced Superconducting Tokamak (EAST) power supply system and analyzes principles of the finite-state machine (FSM) elaborately. Depending on the characteristics of EAST power supply system and the requirements of it, this paper chooses the real-time Linux as system platform and experimental physics and industrial control system (EPICS) as development tool to design a control system of the EAST poloidal field power system supply. This paper utilizes a temporal logic control based on the FSM in order to regulate the operation process of real-time sequence logic control of 12 sets of power controllers. This design is evaluated and tested, and it completely meets the real-time requirement of the poloidal field power supply (PFPS) control system. Several experiments have proved that this design definitely meets the real-time sequence logic control requirement of the PFPS control system in the complex situations under magnetic fields. [ABSTRACT FROM AUTHOR]

The International Thermonuclear Experimental Reactor (ITER) poloidal field (PF) ac/dc converter operates in four quadrants to provide the control of current necessary to PF coils for plasma initialization and the plasma shape control. The technical details for thyristor triggering of high-power ac/dc converter lay the foundation for the regular operation of ITER PF ac/dc converter. With the characteristics such as high power, complex operating modes of ac/dc converter, the synchronization, reliability, and high security have to be critically concerned in the pulse distributor design. In this paper, a pulse distributor is designed to convert the optical signal to high current electrical trigger pulse with less than 1- $\mu \text{s}$ rising time, thereby triggering the paralleled thyristors reliably and isolating 20-kV voltage from the converter unit to controller. Furthermore, the signals of crucial status are online acquired and encoded, then transmitted to controller by optical fiber, thus implementing the online status diagnosis to guarantee the safety operation of ITER PF ac/dc converter. The pulse distributor has been experimented in ITER PF ac/dc converter and all functions have been effectively verified. [ABSTRACT FROM AUTHOR]

Power system operations data are sometimes limited in a given space due to system collinearity. As such, the operations data recorded around an operating point of concern may be deficient or isotropically dispersed. Consequently, online sensitivity identification using ordinary regression methods is prone to large errors. In this paper, a locally weighted ridge regression method is proposed to overcome this problem. The norm-2 Tikhonov–Phillips regularization is integrated into the locally weighted linear regression. The integrated algorithm then has the ability to keep the online sensitivity identification stable if data are collinear while also accommodating the nonlinear and time-varying properties of the sensitivities. The mathematical derivation, online tuning, implementation, and practical considerations of the proposed method are presented. Its effectiveness is validated in a simulation system with operations data measured from real power systems. [ABSTRACT FROM AUTHOR]

The increasing amount of data recorded during power system operations and recently developed data-driven methods make online sensitivity identification (SI) a possibility. However, due to the inherent properties of power systems—nonlinearity, time variance, and collinearity—the effective data that carry the sensitivity information are insufficient. Consequently, the online SI information collected with existing methods may result in unexpected estimates. In this paper, a sufficient effective data condition that guarantees the success of online SI is proposed. The inherent properties of power systems and their impacts on this condition are then investigated. A series of metrics to qualify online whether the data meet the condition is put forward to assess the online SI results. A method is also proposed to select the effective data to improve the online computational efficiency. Finally, the findings and methods are validated in an eight-generator 36-node bus system with operations data recorded from actual power systems. [ABSTRACT FROM AUTHOR]

Measurement sensors installed in the smart transmission system can acquire big data for electromechanical dynamics monitoring. The time-series data obtained carry information of instantaneous relationship of system oscillation modes with respect to operating conditions. To extract this information, this paper proposes a parallel processed online supervised learning algorithm called k-nearest neighbors “locally weighted linear regression” (KNN-LWLR), which is an extensive combination of two famous machine-learning algorithms: 1) the KNN learning; and 2) LWLR learning. Its mathematical derivation, implementation, parameter tuning, and application to electromechanical oscillation mode prediction are first described. The proposed algorithm is then validated based on an 8-generator 36-node system with the real operations data. [ABSTRACT FROM PUBLISHER]