TRANSIENT AND STEADY-STATE LOAD PERFORMANCE OF A STAND-ALONE SELF-EXCITED INDUCTION GENERATOR

An original mathematical model of a stand-alone self-excited induction -generator is presented. The model takes into account the nature and v alue of the load impedance, the load power factor, the exciting termin al capacitance and the rotor speed. The interrelationship of the param eters is demonstrated and their effect on generator performance shown. Resonance is the prime cause of excitation, but the magnetising react ance is the significant factor in determining the bandwidth of success ful self-excitation. Saturation reduces the reactance and limits this bandwidth. The paper shows that a selection of capacitance and loading can compensate for saturation effects on the self-excitation process. Sufficient remanence to initiate self-excitation is assumed and it is shown that there is a critical minimum load impedance and a critical minimum value of terminal capacitance required to permit self-excitati on. The critical value of capacitance for self-excitation is shown to be significantly affected by the rotor speed and the load power-factor . A second mathematical model in the form of a lumped-parameter equiva lent circuit is presented. Analysis shows that the value of capacitanc e used for self-excitation and the nature of the load significantly af fect magnetising reactance. Simulation results are applied to a range of induction machines (>5kW) to be used in stand-alone microhydro gene rating systems. A careful selection of exciting capacitance values rel ated to the external load values is shown to give a band of stable ope ration independent of the magnetic saturation of the machine core. The application of the criteria presented in the two models means that th e selection of the induction machine to be used as the generator is no t a critical factor in system design. Provided that the parameters of the selected machine can be determined by relatively simple tests, a s table stand-alone generating system can easily be designed using the c riteria presented. Extrapolation to larger systems is shown to be stra ightforward.