Harmonics and waveform distortion are substantial power quality concerns for power systems with high penetration of renewable energy generation and non-linear loads. Harmonics and interharmonics should be mitigated for efficient and proper operation of power grids, with high harmonic injection from electric loads and devices. In this paper, an intelligent Active Power Filter (APF) based on Adaptive Linear Neuron (Adaline) is proposed which can provide accurate harmonic estimation and proper mitigation in real-time without any prior knowledge about harmonic/interharmonic orders. A novel formulation is derived, and adaptive learning with momentum is used for training. The proposed APF not only inherits the high adaptability of Adaline APFs but also compensates for their drawbacks by updating weights in the presence of unknown frequency orders. Additionally, an electric topology is proposed where one APF is able to mitigate harmonics of nonlinear loads connected to two adjacent Points of Common Coupling (PCCs). For a radial structure, harmonic currents from the grid are completely mitigated while for a ring structure, only harmonics from one side of the grid are mitigated. Highly varying and distorted load scenarios are studied. Comparison of the proposed APF with the state-of-the- art APFs in the literature proves the effectiveness and suitability of the proposed intelligent APF.