Fast interharmonics and harmonics identification using hybrid wavelet algorithms

The paper deals with the logic, implementation and testing of original algorithms that provide a fast identification of the frequency of interharmonic and harmonic disturbances when they pollute a monitored signal which in stationary state is characterized by periodical medium distortions. Unlike Fourier-based techniques, the identification requires for the (inter)harmonic identification only a small number of samples, corresponding to three quarters from a period of the monitored signal. The algorithms use the details vectors corresponding to the last 5 levels from a decomposition tree with 9 levels generated with 2 different original hybrid wavelet-based algorithms (relying on filters of length 4 and 8). The “key-features” of the details vectors are combined such as to form “polluting fingerprints”, conveniently stored in a matrix along with the limits of ranges of (sub)harmonic orders responsible for their generation. Sets of 5000 tests considering randomly generated polluting (inter)harmonics (covering a frequency range from 150 to 1000 Hz) proved that, when the wavelet-based algorithms were used, the mean runtime consumed for the (inter)harmonic order identification was halved, and the mean and maximum estimation errors were significantly reduced up to values acceptable for practical applications. The modest additional memory and good runtime and errors related performances, along with their relative simple implementation recommend the algorithms as valuable tools in real-time applications for power quality monitoring and fault identification.