Prediction control for a cycloconverter of a power distortion compensation system

Control methods of a flywheel mechanical energy storage system composed of a flywheel-synchronous machine and a current-controlled cycloconverter are described. Improvements to the current control abilities of the cycloconverter, applying modern control and information theories, are discussed. To obtain good current follow characteristics, both deadbeat control and prediction control are applied. The dead-beat scheme possesses one sample delay in its controlled signal which can compensate by predicting reference signals of the next sample time. To predict the load currents at the next sample time, two prediction methods are used. One of the predictions, which has learning ability for repetitive waves using correlation functions of the current, is composed of 256 digital filters. The weighting coefficients of the filters are determined instantaneously to minimize RMS (root mean square) errors. The other prediction, which has a high-frequency response, is applied to compensate random components and high-frequency components. The system is controlled by two digital signal processors for prediction of the currents, and a 16-bit microprocessor for control of the cycloconverter. Experimental results show excellent compensation abilities, not only for repetitive harmonics but also for high-frequency random components