Distortions identification and compensation based on artificial neural networks using symmetrical components of the voltages and the currents

The problem of harmonics identifying and compensating has been of great interest in recent years. A new neural identification scheme for an active power filter (APF) is proposed. This scheme identifies the direct, inverse and zero sequence components of both the voltages and the currents of the power network. The components result from a new and generic decomposition of a three-phase signal which can be either the voltage or the current. For one signal, the direct components extraction requires two independent Adaline networks, and the inverse components extraction two other Adalines. The voltage and current components are used to on-line compute the instantaneous direct, inverse and zero sequence powers. The proposed decomposition is a new formulation of the instantaneous powers and is also appropriate for unbalanced systems. The reference compensation currents can be determined according to different compensation objectives. The resulting compensation currents are then re-injected phase-opposite through the APF in real-time. The performance is evaluated through several simulation examples and through different experiments. The results show that the proposed neural method outperforms other methods, such as the conventional instantaneous power theory.