Removal of measurement noise spikes in grid-connected power converters

Power Electronic Converters (PECs) are routinely used to control the energy supplied to electric grids from distributed generators. The operation of PECs involves the control of voltage and current by rapidly switching power semiconductor devices. Rapid switching of these devices causes a certain class of high frequency noise called spikes which in turn are induced into sensor measurements. In addition to PEC control functions, the sensor measurements are also used to detect over current (OI) over voltage (OV) and under voltage (UV) faults. The presence of noise spikes compromises the performance and leads to unnecessary triggering of fault protection circuitry as a result of false detection of over/under voltage and current faults. The development of a new algorithm to remove short time electrical spikes from PEC sensors is the focus of this paper. In such a way, the PEC is unaffected and false OI, OV, and UV alarms are eliminated. The methodology uses the discrete wavelet transform (DWT) to generate a feature space where the spike feature is detected and removed, and the inverse DWT to reconstruct spike-free measurements. A Field-Programmable Gate Array (FPGA)-based interface board is designed to implement the new algorithm within an actual PEC system. The designed FPGA board along with the new algorithm is used in integration of distributed generators into power systems.