Probabilistic Harmonic Estimation in Uncertain Transmission Networks Using Sequential ANNs

Abstract

In the face of transition towards the decarbonisation, increasing penetration level of power electronic interfaced renewable connections such as wind farms and PV plants are constantly influencing the uncertainties of transmission network and leading to additional uncontrolled harmonic power flows. These potential harmonic distortion issues could result in significant financial losses. To address this problem, the estimation of harmonic propagation through transmission network with increasing penetration of nonlinear loads, power electronics based renewable generation and control devices is becoming increasingly important. This paper proposes a comprehensive framework of applying sequential artificial neural network (ANN) techniques to estimate individual order harmonic distortions and total harmonic distortions (THD) at unmonitored buses in large uncertain transmission networks based on offline measurements and simulations. This study will contribute to facilitate the standard compliance, reduce the extent of the monitor installation, accelerate the assessment of harmonic performance and mitigation studies, as well as contribute to the forecast of potential harmonic issues in large transmission system.

Accuracy of Probabilistic Harmonic Estimation in Sparsely Monitored Transmission Networks

“…It has been trained using individual harmonic distortions of previous weeks at the n th unmonitored bus as the input, and the THD values of previous weeks at the n th unmonitored bus as the target. The advantage of using a sequential ANNs method to estimate THD compared to more traditional approach that uses individual harmonics and conventional formula for calculating THD at unmonitored buses has been clearly demonstrated in [11]. Fig.…”

“…In this study, a sequential ANNs method, introduced in [11], is utilised for the estimation of harmonic distortions in a transmission network, where limited number of harmonic monitoring devices are installed and limited number of harmonic orders are measured. The method combines the successive application of two different two-layer feedforward neural networks namely ANN1 and ANN2.…”

“…Since the accuracy of this probabilistic harmonic estimation method has been validated by one-day harmonic measurements in a preliminary study [11], in this case, the time span for collecting samples has been expanded to one or more weeks, and the period of THD estimation has been extended to one week at the same time.…”

“…A sequential ANNs method [11] has been proposed for reliable estimation of probabilistic harmonic distortions at unmonitored buses in large uncertain transmission networks. As an extension study, this paper evaluates the effectiveness of the sequential ANNs and presents further comprehensive accuracy assessment considering the effect of different types and length of harmonic measurements utilised for training, the effect of amplifying stop errors to optimise training time and the effect of decreasing the number of installed power quality monitors/monitored buses in the large transmission system due to the financial and technical reasons.…”

Propagation characteristics and quantification analysis of harmonic current in large‐scale wind farms

“…Liu et al 25 propose an analytical geometric model of the ellipse formula to analyze the harmonic space propagation and distribution characteristics of uniform lines and the relationship between the weak position of line harmonics as well as the busbar harmonic angle difference of voltage and current through the proposed model. Zhao et al 26 present a sequential artificial neural network method for estimating probabilistic harmonic distortions at unmonitored buses in large uncertain transmission networks. The accuracy of harmonic estimation is affected by the number of installed monitoring equipment, and as a result, the harmonic state of the whole system cannot be accurately estimated.…”