Alternator Identification Using Tailored Multifrequency Signals

This work features alternator response estimation based on spectral analysis techniques using a multifrequency periodic perturbation signal during normal operation. Block diagrams derived from 2-axis theory show the effective closed-loop configuration, which may lead to bias in the spectral estimates. Use of an external periodic perturbation signal and a coherent analysis affords reduced bias and can be used to reduce the effects of incoherent noise on estimates and give a degree of control over the signal to noise ratio. It is shown that of various periodic perturbation signals, based on some performance indicators, multifrequency signals are most suited for frequency domain identification. _ The choice of multifrequency signals is based on comparison of such signals for maintaining low relative peak factors at various parts of the plant. The comparisons were performed by simulation and also on a scaled down model of the alternator and voltage regulator combination. The ability to tailor the spectral envelope of multifrequency signals is explored and a simple solution is provided which overcomes the usual signal deterioration after passage through the intermediary system (automatic voltage regulator) and results in much improved estimates. Identification trials were conducted on a 7.5 kVA microalternator system. Interfacing of the measurement system to a personal computer was carried out. Real-time software was developed which provided automation of data acquisition, storing and processing of measurements. Frequency domain computational results are provided for both unloaded and loaded configurations. For the loaded configuration the relationship between the responses at different operating points is explored.