Undersampled on-line ANN-EKF based estimation of harmonics/interharmonics in power systems

Abstract

This paper describes a new approach to estimate harmonics/interharmonics of power system voltages and currents based on a hybrid artificial neural network (ANN) and extended Kalman filter (EKF) structure. A very low sampling rate is used to implement the proposed ANN-EKF structure with the modest hardware demands. The ANN structure is used for spectral estimation, consumes a very low processing power, and needs a few samples per cycle for real-time implementation. An EKF performs a secondary process on the ANN fundamental frequency phasor estimation to recover it from the errors due to the ANN internal setting. The presented ANN-EKF structure is a very simple and intelligent module to be used by utilities for statistical assessment of harmonics/interharmonics pollution. Besides, it can be a part of utility practices to control, compress, and process large volume of information flowing into data centers through sensors to be placed across the future intelligent power grids. The performance of the proposed method is evaluated by simulations in MATLAB-Simulink environment and confirmed by experimental results.Index Terms-Adaptive Kalman filtering, neural network applications, power system harmonics/interharmonics, signal reconstruction, signal sampling.

Interharmonics in PV systems: a review of analysis and estimation methods; considerations for selection of an apt method

“…Grouping of proposed methodologies parametric/model-based methods spectral model optimisation with high-resolution windowing [15] general noise resilient technique based on MPM [16] robust and adaptive detection in active distribution networks [17] SW-ESPRIT [18] Prony's method [19,20] linear LS method and SVD [21] AR model and Burg algorithm [22] Kalman filtering [23,24] Pisarenko's method [25] min-norm method [26] multiple signal classification [24] subspace-based methods [24] slow sampling and modified gradient search algorithm [27] non-parametric/DFT-based advanced methods DP technique [28] time-domain averaging and DF [29] DFT-based recursive group-harmonic energy distribution [30] interpolating windowed FFT algorithm [31] adaptive window width for interharmonic estimation [32] iterative weighted average phasor method [32] leakage estimation methods [33] synthetic resampling method [34] self-tuning algorithm for harmonic and interharmonic estimations [35] spectral correction method [36] statistical techniques single-channel ICA [37,38] maximum-likelihood estimators [39] SVM algorithm [40] AR model and Burg algorithm [22] machine learning methods SVM algorithm [40] artificial neural network [41,42] generalised optimisation methods PSO [43] FB-based methods filter-based methods with PLL for synchronisation • PLL-based multirate structure for interharmonic estimation [44] • Digital PLL and notch filter for interharmonic estimation…”

A New Fast and Efficient Artificial Neural Network Based State Estimator Incorporated into a Linear Optimal Regulator for Power System Control Enhancement

“…This article presents a new application of state estimation using the feed-forward NN (FFNN) in designing the LOR since calculating Jacobean matrices in the EKF with a conventional method can be a very difficult task, and these equations frequently produce many pages of dense algebra [10]. There has been no similarity with this specific concept, and there are some usages of the concept in other problems, such as in [19], where state estimation is addressed for non-linear discrete-time systems using NNs, or [20], where a new approach to estimate harmonics/inter harmonics of power system voltages and currents based on a hybrid ANN and EKF structure is described. Moreover, the authors in [19] presented estimating a normalized power system transient stability based on the FFNN, and an approach for assessing power system voltage stability based on the FFNN was also presented in the literature.…”