ABSTRACT: A comprehensive system identification procedure is developed to estimate and track the stability parameters of a synchronous generator from time-domain on-line response data. In order to accurately estimate the parameters of the machine model at various test and operating conditions, a multistage identification procedure is proposed. This procedure is verified experimentally by performing both small- and large-disturbance tests on a 7.5 kVA, 220 V, 1800 rpm synchronous generator.
Small disturbance tests are conducted by perturbing the excitation reference voltage in the range 2% to 5% with the generator on-line and delivering power to the infinite bus. Test responses obtained through small disturbance tests are used to estimate stator circuit parameters, field-to-stator turns-ratio, and the field resistance over various operating conditions. Also, by estimating the mutual inductances over several different operating conditions, it is possible to model machine saturation. In this study, artificial neural networks are used to establish generalized saturation models which can take into account several machine variables that are known to influence saturation.
Large disturbance tests are conducted by perturbing the excitation reference voltage in the range 17% to 25% with the generator on-line and delivering power to the infinite bus. Such tests are used to identify rotor body parameters. Large disturbance tests when conducted over a wide range of operating conditions can be used to investigate variations in rotor body parameters as the operating point shifts. Indeed, tests conducted on the 7.5 kVA generator reveal that certain
rotor body parameters are non-linear functions of generator operating condition. A novel artificial neural network based technique is proposed to map variables representative of generator operating condition to each non-linear rotor body parameter.
Artificial neural networks provide a viable means for estimating unmeasurable components of a synchronous generator's state vector by processing sequences of available measurements. These unmeasurable components are typically composed of currents in the rotor body circuits. The proposed observers should account for model parameter non-linearities and provide accurate estimates of rotor body currents irrespective of generator operating condition. By reconstructing the state vector, recursive estimation techniques may be used to estimate machine model parameters.
If your company is a member of the Mechatronic Laboratory, please send the request to receive a copy of any technical report. If you are not a member please send a request to Ali Keyhani, Department of Electrical Engineering, Mechatronics Program at the following address: Ali Keyhani, Ohio State University, Electrical Engineering Department, Mechatronics Systems Laboratory, 2015 Neil Ave., 205 Dereese Lab., Columbus, OH 43210.