A Virtual Array Extension Method for Multiple Partial Discharge Sources Localization Based on Fourth-order Cumulant and Signal Sparse Decomposition

The localization of multiple partial discharges based on UHF signal is difficult to accomplish due to the excessive localization error. A virtual array extension method based on four-order cumulant and signal sparse decomposition is proposed in this paper. The virtually extended array has higher directional resolution and it could detect partial discharge sources with higher accuracy. The 2×2 UHF array that commonly used in existed partial discharge localization methods is virtually extended to 3×3 and 5×5 array. The virtual array extension is accomplished by extending the corresponding steering vector matrix of PD signal spatial spectrum to a overcomplete dictionary, which is then used to realize the sparse decomposition of the PD signal. Eventually the DOA estimation of multiple partial discharge sources is achieved. Simulation and field test results show that the virtually extended UHF array can detect and locate two PD sources simultaneously. The directional angel estimation error is about 6° and the pitch angle error is lower than 4°, which has proved the feasibility of the proposed method.