Due to the rise of electronics and electrical devices into the electrical grid, the network’s power quality has become sensitive. Harmonics parameters in emerging power systems have to be controlled in order to give reliable power/energy to the grid. Adaptive power filter is one of the reliable solutions to compensate harmonics and the unwanted components of the power system. For this effect, we have designed a control strategy for a balanced load in a three-phase four-wire system with an adaptive neural network algorithm. The results obtained for the current harmonic compensation demonstrate a reduction of the total current harmonic distortion below 5%. Those include the reduction in the neutral current and the equilibrium phase currents.