Active power filter (APF) performance is affected by the system delay introduced in the reference signals and/or the actual injected current. This study introduces an artificial intelligent-based technique using a neural network control strategy for a shunt APF. The proposed system is capable of mitigating specific harmonic orders and consequently achieves low harmonic factor and reduced current total harmonic distortion (THD). In addition, the delay time introduced when extracting the harmonic reference current and when controlling the filter current is minimised. The control technique presented in this study is assessed on a medium voltage 6-pulse current source converter where the mains current dominant harmonics (5th, 7th, 11th and 13th) are compensated. The proposed system is simulated using MATLAB/SIMULINK. A prototype system is used to experimentally validate the proposed APF performance.