An Intelligent Technique for Harmonic State Estimation Using Limited Number of Meters

Harmonic monitoring is important in complying with power quality standards and mitigating waveform distortion. A well-developed harmonic monitoring system can provide full observability with limited number of meters. Limiting the number of harmonic meters is equivalent to reducing the harmonic monitoring cost. In this paper, harmonic voltage estimation is addressed by using a limited number of meters which forms an underdetermined system of equations. By employing Artificial Neural Networks (ANN) and Genetic Algorithm (GA), a monitoring system is proposed which outperforms other techniques in terms of accuracy. Two types of ANNs including Deep Convolutional Neural Networks (CNNs) and Feed Forward Neural Network (FFNN) are respectively used to locate harmonic sources and develop pseudo monitoring stations which accordingly will estimate harmonic voltages. GA is employed to guarantee the accuracy of pseudo-estimators. Low frequency harmonics of 3 ^rd , 5 ^th and 7 ^th order have been studied. The proposed method has been simulated for the IEEE-14 bus test systems. The comparison of the proposed methods with state-of-the-art techniques shows the same or improved effectiveness and accuracy, even with a reduced number of measurements.