A New Convolutional Network Structure for Power Quality Disturbance Identification and Classification in Micro-Grids

Abstract

Aiming at the problems of a low convergence speed, low accuracy and poor generalization ability of traditional power disturbance identification and classification methods, a new deep convolutional network structure is presented, and a power quality disturbance identification and classification method for microgrids based on the new network structure is proposed. The network consists of a five-layer one-dimensional modified Inception-residual network (ResNet) (1D-MIR) and a three-layer full-connection tier, which is a deep convolutional network. The idea of the method can be described as follows: First, power disturbance signals are expressed by an n-dimensional unit vector, and a database of these power disturbance signals is established. Second, the disturbance signals in the database are randomly sampled, and the power quality disturbances are calibrated with the n-dimensional unit vector to form both data and test sets. Finally, the gradient descent method and the adaptive moment estimation method (Adam) are adopted to train and optimize the network, respectively, and the trained and optimized network is applied to power quality disturbance identification and classification. A large number of experiments has been conducted, and the obtained results show that the constructed network can quickly extract the characteristics of the various disturbance signals, including single and composite disturbances, and identify and classify them. A comparison of the results obtained by the proposed method with those obtained by several other methods reveals that the proposed method attains a higher accuracy, higher convergence speed and stronger generalization ability. INDEX TERMS New network structure, power quality disturbance detection, deep convolution neural network, deep learning.

On recognition of multiple power quality disturbance events in electricity networks

“…In particular, the classification accuracy of the proposed method can reach 98% which is better than that of the existing PQD detection techniques [5, 7 9]. Deep convolution network [23] also attained high classification accuracy. However, with the increase of PQD classes, its implementation can become more complex.…”

A Framework Based on Machine Learning for Analytics of Voltage Quality Disturbances

“…For this case, the notation means that the angle ϕ is between 0 and 30 degrees, including 30 degrees; i.e., 0 < ϕ ≤ 30. (0, 30] I b between phase "b" and neutral Phase "b" (30,60] I I c between phase "a" and "b" Phases "a" and "b" (60, 90] I a between phase "a" and neutral Phase "a" (90, 120] I I b between phase "a" and "c" Phases "a" and "c" (120,150] I c between phase "c" and neutral Phases "c" (150, 180] I I a between phase "b" and "c" Phases "b" and "c"…”

“…Diverse DL techniques were applied to power quality data such as convolutional neural networks (CNNs), longshort-term memory (LSTM), generative adversarial networks (GANs), and deep autoencoder (DAE). CNNs are applied for the classification of PQ disturbances as in [27][28][29][30][31], voltage dip classification [5], recognition of voltage dip causes [32], and prediction of harmonics [33][34][35]. LSTM is applied to classification of events [36,37], recognition of voltage dip causes [38], voltage dip classification [39], and harmonic prediction [33].…”

Comprehensive Review on Detection and Classification of Power Quality Disturbances in Utility Grid With Renewable Energy Penetration

“…the computational burden in both noisy and non-noisy conditions of real-time operation [169], [217]. However, better identification of PQDs has been found with convolutional network structure in different noise levels [218]. Multifusion convolutional neural network for complex PQDs in the noisy environment has been presented in [219].…”