Sliding reinforced competitive learning scheme for voltage and frequency regulation of diesel engine driven standalone single-phase generators

摘要

This paper presents the sliding reinforced competitive learning (SR-CL) algorithm for power quality improvement and its implementation for voltage and frequency regulation of diesel engine driven standalone single phase self excited induction generator (SEIG) using digital signal processor (DSP) and a two leg voltage source converter (VSC) bridge. The proposed control algorithm iterates continuously in real time through the system input variables (training data). The proposed SR-CL algorithm updates the reference generator current vector in response to every successive sampling of input variables. The source current, source voltage, load current and frequency are the input variables of the control scheme. The information about the revised instantaneous values of these quantities is desired by the algorithm in every sampling period. The learning rate of the proposed SR-CL neural network algorithm decays to zero when the error between the actual generator current and reference generator current reduces to prescribed minimum level of error after few iterations. The algorithm requires less computation time and memory since no data is required to be stored in the memory during real time implementation of the scheme. The proposed control is flexible and stable enough to handle all types of non linear relationship among frequency, voltage, magnetizing reactance and speed of the generator to obtain the optimum system efficiency during all types of loading conditions. The detailed design of the system and experimental validation of the scheme are also presented in this paper.