# Power Disturbance Monitoring through Techniques for Novelty Detection on Wind Power and Photovoltaic Generation

^{1}

^{2}

^{3}

^{*}

## Abstract

**:**

## 1. Introduction

- Probabilistic-based. Here, the data distribution can be thresholded, which in turn can be useful for defining in the data space limits or boundaries of normality. With this information, it is possible to determine if a single sample belongs to the same distribution or not. For such a task, it would be necessary to estimate probability density functions, for example, Gaussian mixture model (GMM) [14].
- Domain-based. These approaches need the generation or definition of boundaries according to the form of the training data set. It means that they describe the domain of target classes being insensible to the sampling and density, such as one-class support vector machines (OC-SVM) [17].
- Reconstruction-based. These methodologies are widely used in applications that require classification tasks and regression purposes. Normally, these techniques perform data modeling automatically without supervision, and the estimation or prediction is used to obtain a performance metric. That metric is the difference between a test vector and an output vector, better known as the reconstruction error. Thus, this metric is related as a novelty score. For example, there are stacked auto-encoders (SAE) [18] and self-organizing maps (SOM) [19].
- Theoretic information-based. For these approaches, the entire data set is used for computing specific metrics, such as entropies, energies, forms, and moments. With this information, the novelty means alterations in these values compared with normal data sets, such as time–frequency ridge estimation (TFRE) [20,21], degree of cyclostationarity (DCS) demodulation [22], and entropy measure-based methods (EMBM) [23].

## 2. Theoretical Foundations

#### 2.1. Stacked Autoencoder (SAE)

_{(x)}, as expressed in (1).

#### 2.2. One-Class Support Vector Machine

#### 2.3. k-Nearest Neighbors

#### 2.4. Gaussian Mixtrure Model

- A mean $\mu $ defining its center.
- A covariance $\mathsf{\Sigma}$ specifying its width.
- A mixture probability $\pi $ that defines how large or small the Gaussian function will be.

#### 2.5. Self-Organizing Maps

- Nearby learning nodes have nearby weight vectors.
- Each input data has some close representative among the learning nodes (in the sense that there is some learning node whose weight vector resembles the input data).

#### 2.6. Isolation Forest

## 3. Proposed Methodology

## 4. Experimental Results

#### 4.1. Experimental Setup

#### 4.2. Results of the Signals from the SPG_DS

#### 4.2.1. SPG_DS 1

#### 4.2.2. SPG_DS 2

#### 4.2.3. SPG_DS 3

#### 4.2.4. SPG_DS 4

#### 4.2.5. SPG_DS 5

#### 4.3. Analysis of the Signals from the WPG_DS

#### 4.3.1. WPG_DS 1

#### 4.3.2. WPG_DS 2

#### 4.3.3. WPG_DS 3

#### 4.3.4. WPG_DS 4

#### 4.3.5. WPG_DS 5

## 5. Discussion

#### 5.1. Discussion of the Results from the SPG_DS

#### 5.1.1. SPG_DS 1

#### 5.1.2. SPG_DS 2

#### 5.1.3. SPG_DS 3

#### 5.1.4. SPG_DS 4

#### 5.1.5. SPG_DS 5

#### 5.2. Discussion of the Results from the WPG_DS

#### 5.2.1. WPG_DS 1

#### 5.2.2. WPG_DS 2

#### 5.2.3. WPG_DS 3

#### 5.2.4. WPG_DS 4

#### 5.2.5. WPG_DS 5

## 6. Conclusions

## Author Contributions

## Funding

## Data Availability Statement

## Acknowledgments

## Conflicts of Interest

## Appendix A

Indicator | Equation | |
---|---|---|

Mean | $\overline{x}=\frac{1}{n}\xb7{\displaystyle \sum}_{i=1}^{n}({x}_{i})$ | (A1) |

Maximum Value | $\widehat{x}=\mathrm{max}\left(x\right)$ | (A2) |

Root Mean Square (RMS) | $RMS=\sqrt{\frac{1}{n}\xb7{\displaystyle \sum}_{i=1}^{n}{({x}_{i})}^{2}}$ | (A3) |

Square Root Mean (SRM) | $SRM={\left(\frac{1}{n}\xb7{\displaystyle \sum}_{i=1}^{n}\sqrt{\left|({x}_{i}\right|}\right)}^{2}$ | (A4) |

Standard Deviation | $\sigma =\sqrt{\frac{1}{n}\xb7{\displaystyle \sum}_{i=1}^{n}{({x}_{i}-\overline{x})}^{2}}$ | (A5) |

Variance | ${\sigma}^{2}=\frac{1}{n}\xb7{\displaystyle \sum}_{i=1}^{n}{({x}_{i}-\overline{x})}^{2}$ | (A6) |

RMS with shape factor | $S{F}_{RMS}=\frac{RMS}{\frac{1}{n}\xb7{{\displaystyle \sum}}_{i=1}^{n}\left|{x}_{i}\right|}$ | (A7) |

SRM with shape factor | $S{F}_{SRM}=\frac{SRM}{\frac{1}{n}\xb7{{\displaystyle \sum}}_{i=1}^{n}\left|{x}_{i}\right|}$ | (A8) |

Crest Factor | $CF=\frac{\widehat{x}}{RMS}$ | (A9) |

Latitude Factor | $LF=\frac{\widehat{x}}{SRM}$ | (A10) |

Impulse Factor | $IF=\frac{\widehat{x}}{\frac{1}{n}\xb7{{\displaystyle \sum}}_{i=1}^{n}\left|{x}_{i}\right|}$ | (A11) |

Skewness | ${S}_{k}=\frac{E\left[{({x}_{i}-\overline{x})}^{3}\right]}{{\sigma}^{3}}$ | (A12) |

Kurtosis | $k=\frac{E\left[{({x}_{i}-\overline{x})}^{4}\right]}{{\sigma}^{4}}$ | (A13) |

5th Moment | $5t{h}_{M}=\frac{E\left[{({x}_{i}-\overline{x})}^{5}\right]}{{\sigma}^{5}}$ | (A14) |

6th Moment | $6t{h}_{M}=\frac{E\left[{({x}_{i}-\overline{x})}^{6}\right]}{{\sigma}^{6}}$ | (A15) |

Energy | $E={\displaystyle \sum}_{i=1}^{n}\left({\left|{x}_{i}\right|}^{2}\right)$ | (A16) |

Entropy | $ET=-{\displaystyle \sum}_{i=1}^{n}{x}_{i}{}^{2}\mathrm{log}\left({x}_{i}{}^{2}\right)$ | (A17) |

Range | $R=Max\left({x}_{i}\right)-Min\left({x}_{i}\right)$ | (A18) |

Form factor | $FF=\frac{\overline{{x}_{i}}}{RM{S}_{i}}$ | (A19) |

Log energy Entropy | $LE=-{\displaystyle \sum}_{i=1}^{n}\mathrm{log}\left({x}_{i}{}^{2}\right)$ | (A20) |

## References

- Chen, G.Q.; Wu, X.D.; Guo, J.; Meng, J.; Li, C. Global Overview for Energy Use of the World Economy: Household-Consumption-Based Accounting Based on the World Input-Output Database (WIOD). Energy Econ.
**2019**, 81, 835–847. [Google Scholar] [CrossRef] - Climate Change and COP26: Are Digital Technologies and Information Management Part of the Problem or the Solution? An Editorial Reflection and Call to Action-ScienceDirect. Available online: https://www.sciencedirect.com/science/article/pii/S0268401221001493?via%3Dihub (accessed on 18 December 2022).
- Global Wind Energy Council (GWEC). Global Wind Report 2022; Global Wind Energy Council: Brussels, Belgium, 2022. [Google Scholar]
- REN21. REN21 Renewables Now. In Renewables 2022 Global Status Report; REN21: Paris, France, 2022. [Google Scholar]
- Zhang, X.-P.; Yan, Z. Energy Quality: A Definition. IEEE Open Access J. Power Energy
**2020**, 7, 430–440. [Google Scholar] [CrossRef] - Power Quality in Microgrids Including Supraharmonics: Issues, Standards, and Mitigations | IEEE Journals & Magazine | IEEE Xplore. Available online: https://ieeexplore.ieee.org/document/9136692 (accessed on 18 December 2022).
- Rodriguez-Guerrero, M.A.; Jaen-Cuellar, A.Y.; Carranza-Lopez-Padilla, R.D.; Osornio-Rios, R.A.; Herrera-Ruiz, G.; Romero-Troncoso, R.d.J. Hybrid Approach Based on GA and PSO for Parameter Estimation of a Full Power Quality Disturbance Parameterized Model. IEEE Trans. Ind. Inform.
**2018**, 14, 1016–1028. [Google Scholar] [CrossRef] - Xiao, F.; Ai, Q. Data-Driven Multi-Hidden Markov Model-Based Power Quality Disturbance Prediction That Incorporates Weather Conditions. IEEE Trans. Power Syst.
**2019**, 34, 402–412. [Google Scholar] [CrossRef] - Liu, S.; Zheng, C.; Zhang, B.; Dai, S.; Tang, Y.; Wang, Y. A Data-Driven Self-Learning Evaluation Method of Voltage Sag Severity. CPSS Trans. Power Electron. Appl.
**2022**, 7, 328–334. [Google Scholar] [CrossRef] - Cui, C.; Duan, Y.; Hu, H.; Wang, L.; Liu, Q. Detection and Classification of Multiple Power Quality Disturbances Using Stockwell Transform and Deep Learning. IEEE Trans. Instrum. Meas.
**2022**, 71, 1–12. [Google Scholar] [CrossRef] - Perera, S.; Elphick, S. Chapter 7-Implications of Equipment Behaviour on Power Quality. In Applied Power Quality; Perera, S., Elphick, S., Eds.; Elsevier: Amsterdam, The Netherlands, 2023; pp. 185–258. ISBN 978-0-323-85467-2. [Google Scholar]
- Ouafae, B.; Oumaima, L.; Mariam, R.; Abdelouahid, L. Novelty Detection Review State of Art and Discussion of New Innovations in The Main Application Domains. In Proceedings of the 2020 1st International Conference on Innovative Research in Applied Science, Engineering and Technology (IRASET), Meknes, Morocco, 16–19 April 2020; pp. 1–7. [Google Scholar]
- Pimentel, M.A.F.; Clifton, D.A.; Clifton, L.; Tarassenko, L. A Review of Novelty Detection. Signal Process.
**2014**, 99, 215–249. [Google Scholar] [CrossRef] - Patel, E.; Kushwaha, D.S. Clustering Cloud Workloads: K-Means vs. Gaussian Mixture Model. Procedia Comput. Sci.
**2020**, 171, 158–167. [Google Scholar] [CrossRef] - Zhang, S. Challenges in KNN Classification. IEEE Trans. Knowl. Data Eng.
**2022**, 34, 4663–4675. [Google Scholar] [CrossRef] - Shao, C.; Du, X.; Yu, J.; Chen, J. Cluster-Based Improved Isolation Forest. Entropy
**2022**, 24, 611. [Google Scholar] [CrossRef] - Xing, H.-J.; Li, L.-F. Robust Least Squares One-Class Support Vector Machine. Pattern Recognit. Lett.
**2020**, 138, 571–578. [Google Scholar] [CrossRef] - Yan, B.; Han, G. Effective Feature Extraction via Stacked Sparse Autoencoder to Improve Intrusion Detection System. IEEE Access
**2018**, 6, 41238–41248. [Google Scholar] [CrossRef] - Valverde Castilla, G.A.; Mira McWilliams, J.M.; González-Pérez, B. One-Layer vs. Two-Layer SOM in the Context of Outlier Identification: A Simulation Study. Appl. Sci.
**2021**, 11, 6241. [Google Scholar] [CrossRef] - Li, Y.; Zhang, X.; Chen, Z.; Yang, Y.; Geng, C.; Zuo, M.J. Time-Frequency Ridge Estimation: An Effective Tool for Gear and Bearing Fault Diagnosis at Time-Varying Speeds. Mech. Syst. Signal Process.
**2023**, 189, 110108. [Google Scholar] [CrossRef] - Li, Y.; Yang, Y.; Feng, K.; Zuo, M.J.; Chen, Z. Automated and Adaptive Ridge Extraction for Rotating Machinery Fault Detection. IEEE/ASME Trans. Mechatron.
**2023**, 1–11. [Google Scholar] [CrossRef] - Zhen, D.; Li, D.; Feng, G.; Zhang, H.; Gu, F. Rolling Bearing Fault Diagnosis Based on VMD Reconstruction and DCS Demodulation. IJHM
**2022**, 5, 205. [Google Scholar] [CrossRef] - Zhou, Y.; Kumar, A.; Parkash, C.; Vashishtha, G.; Tang, H.; Glowacz, A.; Dong, A.; Xiang, J. Development of Entropy Measure for Selecting Highly Sensitive WPT Band to Identify Defective Components of an Axial Piston Pump. Appl. Acoust.
**2023**, 203, 109225. [Google Scholar] [CrossRef] - Xie, C.-H.; Chang, J.-Y.; Liu, Y.-J. Estimating the Number of Components in Gaussian Mixture Models Adaptively for Medical Image. Optik
**2013**, 124, 6216–6221. [Google Scholar] [CrossRef] - Gournelos, T.; Kotinas, V.; Poulos, S. Fitting a Gaussian Mixture Model to Bivariate Distributions of Monthly River Flows and Suspended Sediments. J. Hydrol.
**2020**, 590, 125166. [Google Scholar] [CrossRef] - Sarmadi, H.; Karamodin, A. A Novel Anomaly Detection Method Based on Adaptive Mahalanobis-Squared Distance and One-Class KNN Rule for Structural Health Monitoring under Environmental Effects. Mech. Syst. Signal Process.
**2020**, 140, 106495. [Google Scholar] [CrossRef] - Mesarcik, M.; Ranguelova, E.; Boonstra, A.-J.; van Nieuwpoort, R.V. Improving Novelty Detection Using the Reconstructions of Nearest Neighbours. Array
**2022**, 14, 100182. [Google Scholar] [CrossRef] - Susto, G.A.; Beghi, A.; McLoone, S. Anomaly Detection through On-Line Isolation Forest: An Application to Plasma Etching. In Proceedings of the 2017 28th Annual SEMI Advanced Semiconductor Manufacturing Conference (ASMC), Saratoga Springs, NY, USA, 15–18 May 2017; pp. 89–94. [Google Scholar]
- Wang, Y.; Li, X. An Innovative Huffman Forest-Based Method to Detected Railroad Station Anomalies. Sensors
**2022**, 22, 3915. [Google Scholar] [CrossRef] [PubMed] - Catania, C.A.; Bromberg, F.; Garino, C.G. An Autonomous Labeling Approach to Support Vector Machines Algorithms for Network Traffic Anomaly Detection. Expert Syst. Appl.
**2012**, 39, 1822–1829. [Google Scholar] [CrossRef] - Beghi, A.; Cecchinato, L.; Corazzol, C.; Rampazzo, M.; Simmini, F.; Susto, G.A. A One-Class SVM Based Tool for Machine Learning Novelty Detection in HVAC Chiller Systems. IFAC Proc. Vol.
**2014**, 47, 1953–1958. [Google Scholar] [CrossRef] - Sun, Z.; Sun, H. Stacked Denoising Autoencoder with Density-Grid Based Clustering Method for Detecting Outlier of Wind Turbine Components. IEEE Access
**2019**, 7, 13078–13091. [Google Scholar] [CrossRef] - Ou, J.; Li, H.; Huang, G.; Zhou, Q. A Novel Order Analysis and Stacked Sparse Auto-Encoder Feature Learning Method for Milling Tool Wear Condition Monitoring. Sensors
**2020**, 20, 2878. [Google Scholar] [CrossRef] - Saucedo-Dorantes, J.J.; Delgado-Prieto, M.; Osornio-Rios, R.A.; Romero-Troncoso, R.d.J. Industrial Data-Driven Monitoring Based on Incremental Learning Applied to the Detection of Novel Faults. IEEE Trans. Ind. Inform.
**2020**, 16, 5985–5995. [Google Scholar] [CrossRef] - Li, M.; Kashef, R.; Ibrahim, A. Multi-Level Clustering-Based Outlier’s Detection (MCOD) Using Self-Organizing Maps. Big Data Cogn. Comput.
**2020**, 4, 24. [Google Scholar] [CrossRef] - Jaen-Cuellar, A.Y.; Morales-Velazquez, L.; Romero-Troncoso, R.d.J.; Moriñigo-Sotelo, D.; Osornio-Rios, R.A. Micro-Genetic Algorithms for Detecting and Classifying Electric Power Disturbances. Neural Comput. Appl.
**2017**, 28, 379–392. [Google Scholar] [CrossRef] - Perera, S.; Elphick, S. Chapter 1-Introduction to Power Quality in Modern Power Systems. In Applied Power Quality; Perera, S., Elphick, S., Eds.; Elsevier: Amsterdam, The Netherlands, 2023; pp. 1–17. ISBN 978-0-323-85467-2. [Google Scholar]
- IEEE Std 1159–1995; IEEE Recommended Practice for Monitoring Electric Power Quality. IEEE: Piscataway, NJ, USA, 1995; pp. 1–80. [CrossRef]
- International Standard IEC 61000-4-30; Electromagnetic Compatibility (EMC)-Part 4-30: Testing and Measurement Techniques-Power Quality Measurement Methods. IEEE: Piscataway, NJ, USA, 2008.
- Wu, T.; Angela Zhang, Y.-J.; Tang, X. Online Detection of Events with Low-Quality Synchrophasor Measurements Based on $i$Forest. IEEE Trans. Ind. Inform.
**2021**, 17, 168–178. [Google Scholar] [CrossRef] - Zhu, R.; Gong, X.; Hu, S.; Wang, Y. Power Quality Disturbances Classification via Fully-Convolutional Siamese Network and k-Nearest Neighbor. Energies
**2019**, 12, 4732. [Google Scholar] [CrossRef] - Karasu, S.; Saraç, Z. Classification of Power Quality Disturbances by 2D-Riesz Transform, Multi-Objective Grey Wolf Optimizer and Machine Learning Methods. Digit. Signal Process.
**2020**, 101, 102711. [Google Scholar] [CrossRef] - Thirumala, K.; Pal, S.; Jain, T.; Umarikar, A.C. A Classification Method for Multiple Power Quality Disturbances Using EWT Based Adaptive Filtering and Multiclass SVM. Neurocomputing
**2019**, 334, 265–274. [Google Scholar] [CrossRef] - Kapoor, R.; Gupta, R.; Son, L.H.; Jha, S.; Kumar, R. Detection of Power Quality Event Using Histogram of Oriented Gradients and Support Vector Machine. Measurement
**2018**, 120, 52–75. [Google Scholar] [CrossRef] - Fu, L.; Zhu, T.; Pan, G.; Chen, S.; Zhong, Q.; Wei, Y. Power Quality Disturbance Recognition Using VMD-Based Feature Extraction and Heuristic Feature Selection. Appl. Sci.
**2019**, 9, 4901. [Google Scholar] [CrossRef] - Elvira-Ortiz, D.A.; Saucedo-Dorantes, J.J.; Osornio-Rios, R.A.; Morinigo-Sotelo, D.; Antonino-Daviu, J.A. Power Quality Monitoring Strategy Based on an Optimized Multi-Domain Feature Selection for the Detection and Classification of Disturbances in Wind Generators. Electronics
**2022**, 11, 287. [Google Scholar] [CrossRef] - Jaen-Cuellar, A.Y.; Osornio-Ríos, R.A.; Trejo-Hernández, M.; Zamudio-Ramírez, I.; Díaz-Saldaña, G.; Pacheco-Guerrero, J.P.; Antonino-Daviu, J.A. System for Tool-Wear Condition Monitoring in CNC Machines under Variations of Cutting Parameter Based on Fusion Stray Flux-Current Processing. Sensors
**2021**, 21, 8431. [Google Scholar] [CrossRef]

Category | Technique for ND | Ref. | Application |
---|---|---|---|

Probabilistic based | GMM | [24] | Methodology for accurately estimating and testing the number of components in medical images, by model selection criterion, where the image can be, for example, a tomography. It is also defined as a sum of weighted real parts of all log-characteristic functions of the GMM as new convergent function, and it is validated for simulated datasets and medical images in two dimensions of univariate acidity. |

[25] | Estimation of the monthly river flows and their suspended sediments, by using the sample points of the rivers, in such a way that their parameters are adjusted by comparisons of information criteria. The parameters tuned, also known as latent variables, are associated with a mechanism that generates bivariate distributions between two regimes related to the seasons of the hydrological variables. The transitions between the regimes are estimated through the Markov chain, which is a switched regression model that finally predicts the monthly suspended sediment fluxes when the water discharge is known. | ||

Distance based | kNN | [26] | Development of an unsupervised learning strategy for monitoring health under environmental variability based on the combination of the Mahalanobis-squared distance and the one-class kNN rule. This approach trains and tests the datasets by finding as many neighbors as possible, removing the environmental variability conditions, estimating the covariance matrices, and using a generalized model of extreme values distribution. |

[27] | Development of a semi-supervised method of novelty detection in datasets with different characteristics through the reconstruction of the nearest neighbors and the latent-neighbor distances of a given input. | ||

IF | [28] | Detection of anomalies (outliers and inliers) in data acquired from the manufacturing process of etching. The proposed approach was divided into two main parts: offline stage, for evaluating performance and for updating the approach; online stage, for evaluating the observations at hand. | |

[29] | Methodology for detecting railroad station anomalies based on an improvement of IF, Huffman forest. This scheme leverages Huffman encoding to measure abnormalities in diverse railroad scenarios. Thus, the trees of the forest are integrated from the perspective of data points for computing anomalies scores of the instances, considering the local and the global available information. | ||

Domain based | OC-SVM | [30] | An unsupervised classification algorithm that autonomously labels between anomalies and normal traffic in the network for imbalance in classes distributions. |

[31] | Prediction of faults and detection of unknown status in a heating, ventilation, and air conditioning (HVAC) chiller system. Here, the principal component analysis (PCA) is used for remarking the novelties in the chiller, considering the anomalies as faults, the condenser fouling, the condenser water flowing, and the refrigerant leakage as priori unknown conditions. | ||

Reconstruction based | SAE | [32] | Detection of outlier components in data coming from wind turbines (WT). Thus, an unsupervised approach for detecting outliers is developed by combining the stacked denoising auto-encoder (SDAE), for extracting the features from the original data and training purposes, and a density-grid-based method, used for data clustering. |

[33] | Monitoring of tool wear for the continuous cutting process in a milling machine under variable speed conditions. This task is achieved by using the stacked sparse auto-encoder (SSAE), and the measured signals were converted into angle domain stationary signals through the order analysis for performing feature extraction. | ||

SOM | [34] | Detecting unknown problems in electromechanical systems through multi-fault detection and identification scheme. Three main parts integrate such scheme, the feature extraction in the time domain directly from multiple sensors readings, the data-based model of available operating conditions through SOM, and the incremental learning of new conditions through self-organizing structures. | |

[35] | Detecting outliers in business applications for preventing losses and for optimizing the revenue. This is achieved through a set of self-organizing structures that recognize unusual behaviors on data patterns. | ||

Information theoretic based | TFRE | [20] | Monitoring and diagnosis of faults in gears and bearing of rotating machinery through oscillatory components with time-varying amplitudes and frequencies. The component with a sequence of peaks in the time–frequency representation is known as ridge. These ridges enhance the machine condition assessment but require an adequate cost kernel and an adaptive search region detection principle for ensuring the ridges smoothness. |

[21] | Detection of faults in rotating machinery through an automated and adaptive ridge extraction (AARE). For avoiding interferences when searching a ridge, an adaptive edge detection strategy is implemented. In addition, an adaptive core function constructed by using the signal characteristics keeps stability during peak exploration and curve continuity. This technique runs automatically because it does not require parameter tuning, minimizing user intervention. | ||

DCS | [22] | Fault diagnosis in rolling bearings through the variational mode decomposition (VMD) and degree of cyclostationarity (DCS) demodulation. At first place, the reconstruction of denoised signal is carried out by means of sparsity-based reconstruction factor. Next, fault characteristics are extracted through DCS, and finally, combined with VMD, the faults detection is performed. | |

EMBM | [23] | Identification of defective components, case of the rolling bearings, on axial piston pumps. The fault detection is made through a tangent hyperbolic fuzzy entropy measure-based method that determines the most sensitive Wavelet packet transform (WPT). |

**Table 2.**Fragment of the PQD classification according to IEEE-1159 standard [38].

PQD | Duration | Magnitude | Spectral Content |
---|---|---|---|

Transients—Oscillatory | |||

Low frequency | 0.3–50 ms | 0–4 pu ^{1} | <5 kHz |

Medium frequency | 20 µs | 0–8 pu | 5–500 kHz |

High frequency | 5 µs | 0–4 pu | 0.5–5 MHz |

Short duration variations—Instantaneous | |||

Sag | 0.5–30 cycles | 0.1–0.9 pu | |

Swell | 0.5–30 cycles | 1.1–1.8 pu | |

Short duration variations—Momentary | |||

Sag | 30 cycles–3 s | 0.1–0.9 pu | |

Swell | 30 cycles–3 s | 1.1–1.4 pu | |

Short duration variations—Temporary | |||

Sag | 3 s–1 min | 0.1–0.9 pu | |

Swell | 3 s–1 min | 1.1–1.2 pu | |

Voltage fluctuations | |||

Flicker | intermittent | 0.1–7% | <25 Hz |

^{1}The units are dimensionless; they are considered instead per unit (pu).

**Table 3.**Fragment of the power quality parameters according to IEC-61000-4-30 standard [39].

Parameters | Class | Measurement Method | Uncertainity | Measuring Range |
---|---|---|---|---|

Flicker | A | IEC 61000-4-15 | IEC 61000-4-15 | 0.2~10.0 P_{st} |

S | IEC 61000-4-15 | Twice the permitted measurement uncertainty required by IEC 61000-4-15 | 0.4~4.0 P_{st} | |

Dips and Swells | A | U_{rms(1/2)} | Amplitude ±0.2% U_{din}Duration +/− 1 cycle | N/A |

S | U_{rms}_{(1/2)} on each measurement channel, or U_{rms}_{(1)} on each measurement channel. | Amplitude ±1% of U_{din}Duration +/− 1 cycle or +/− 2 cycles | N/A | |

Transient voltages IEC 61180 | A | N/R | N/R | N/R |

S | N/R | N/R | N/R | |

Fast transients IEC 61000-4-4 | A | N/R | N/R | N/R |

S | N/R | N/R | N/R |

_{din}= value obtained from the declared supply voltage by a transducer ratio, U

_{rms}

_{(1/2)}= value of the r.m.s., voltage measured over 1 cycle, commencing at a fundamental zero crossing, and refreshed each half-cycle, U

_{rms}

_{(1)}= value of the r.m.s., voltage measured over 1 cycle and refreshed each cycle, P

_{st}= short-term flicker severity, unless otherwise specified, the evaluation time is 10 min, Subclasses advanced (A) and surveys (S) are considered, and subclass basic (B) will be removed in future from the standard.

ND Technique | Hyperparameters | Description | Evaluated Values | Best Value |
---|---|---|---|---|

kNN | k | Number of the closest neighbors | 3, 5, 7, 9, 11 | 3 |

OC-SVM | v | Regularization value | 0.025, 0.03, 0.035, 0.04, 0.045 | 0.03 |

GMM | RV | Regularization value | 0.0001, 0.001, 0.01, 0.1 | 0.01 |

SOM | Q_{re} | Percentile for reconstruction error | 90, 95, 98 | 90 |

IF | n | Structure dimension | 8, 10, 12 | 8 |

SAE ^{1} | Q_{u} | Percentile of tree depth threshold for assigning the score of novelty | 90, 95, 98 | 90 |

^{1}For SAE, a structure of 50-30-25 was configured for each case study, through several grid searches for the hyperparameters of L2W regularization, sparsity proportion, sparsity regularization, and epochs.

**Table 5.**Numerical results of the novelty detection obtained for the first case study of the SPG_DS.

ND Technique | Known (%) | Novelty (%) | Total Performance (%) |
---|---|---|---|

SAE | 88 | 100 | 94 |

OC-SVM | 98.5 | 100 | 99.25 |

kNN | 94.5 | 100 | 97.25 |

GMM | 78 | 100 | 89 |

SOM | 96.5 | 100 | 98.25 |

IF | 96 | 100 | 98 |

**Table 6.**Numerical results of the novelty detection obtained for the second case study of the SPG_DS.

ND Technique | Known (%) | Novelty (%) | Total Performance (%) |
---|---|---|---|

SAE | 87 | 100 | 93.5 |

OC-SVM | 98.5 | 100 | 99.25 |

kNN | 94.5 | 100 | 97.25 |

GMM | 81.5 | 100 | 90.75 |

SOM | 95.5 | 100 | 97.75 |

IF | 96 | 100 | 98 |

**Table 7.**Numerical results of the novelty detection obtained for the third case study of the SPG_DS.

ND Technique | Known (%) | Novelty (%) | Total Performance (%) |
---|---|---|---|

SAE | 95 | 100 | 97.5 |

OC-SVM | 79.5 | 100 | 89.75 |

kNN | 93.5 | 95 | 94.25 |

GMM | 66 | 98.5 | 82.25 |

SOM | 99 | 15 | 57 |

IF | 97.5 | 86.5 | 89.5 |

**Table 8.**Numerical results of the novelty detection obtained for the fourth case study of the SPG_DS.

ND Technique | Known (%) | Novelty (%) | Total Performance (%) |
---|---|---|---|

SAE | 90.5 | 100 | 92.25 |

OC-SVM | 83.5 | 100 | 91.75 |

kNN | 91.5 | 100 | 95.75 |

GMM | 77 | 100 | 88.5 |

SOM | 91 | 69 | 77.5 |

IF | 97 | 80 | 88.5 |

**Table 9.**Numerical results of the novelty detection obtained for the fifth case study of the SPG_DS.

ND Technique | Known (%) | Novelty (%) | Total Performance (%) |
---|---|---|---|

SAE | 87.5 | 100 | 93.75 |

OC-SVM | 99 | 100 | 99.5 |

kNN | 94 | 100 | 97 |

GMM | 82.5 | 100 | 91.25 |

SOM | 96.5 | 90 | 93.25 |

IF | 96.5 | 100 | 98.25 |

**Table 10.**Numerical results of the novelty detection obtained for the first case study of the WPG_DS.

ND Technique | Known (%) | Novelty (%) | Total Performance (%) |
---|---|---|---|

SAE | 87.33 | 100 | 93.68 |

OC-SVM | 98.33 | 100 | 99.18 |

kNN | 93.66 | 100 | 96.85 |

GMM | 80.68 | 100 | 90.33 |

SOM | 96 | 75.01 | 85.5 |

IF | 93.66 | 100 | 96.84 |

**Table 11.**Numerical results of the novelty detection obtained for the second case study of the WPG_DS.

ND Technique | Known (%) | Novelty (%) | Total Performance (%) |
---|---|---|---|

SAE | 85.01 | 100 | 92.5 |

OC-SVM | 100 | 100 | 100 |

kNN | 94.33 | 100 | 97.18 |

GMM | 80.33 | 100 | 90.17 |

SOM | 97.33 | 76.67 | 87.01 |

IF | 95.33 | 100 | 97.68 |

**Table 12.**Numerical results of the novelty detection obtained for the third case study of the WPG_DS.

ND Technique | Known (%) | Novelty (%) | Total Performance (%) |
---|---|---|---|

SAE | 87.34 | 100 | 93.66 |

OC-SVM | 69.68 | 100 | 84.83 |

kNN | 91.67 | 100 | 95.83 |

GMM | 74.34 | 98.34 | 86.34 |

SOM | 92.65 | 80.01 | 86.35 |

IF | 94.66 | 81.67 | 88.18 |

**Table 13.**Numerical results of the novelty detection obtained for the fourth case study of the WPG_DS.

ND Technique | Known (%) | Novelty (%) | Total Performance (%) |
---|---|---|---|

SAE | 88 | 100 | 94 |

OC-SVM | 64.33 | 100 | 82.17 |

kNN | 91 | 100 | 95.5 |

GMM | 83.66 | 86.67 | 85.18 |

SOM | 95 | 25.34 | 60.16 |

IF | 96.66 | 47.67 | 72.17 |

**Table 14.**Numerical results of the novelty detection obtained for the fifth case study of the WPG_DS.

ND Technique | Known (%) | Novelty (%) | Total Performance (%) |
---|---|---|---|

SAE | 82.33 | 100 | 91.18 |

OC-SVM | 99.33 | 100 | 99.67 |

kNN | 91.99 | 100 | 96.01 |

GMM | 80.67 | 100 | 90.33 |

SOM | 92.33 | 75.34 | 83.84 |

IF | 93.67 | 100 | 96.84 |

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |

© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Gonzalez-Abreu, A.D.; Osornio-Rios, R.A.; Elvira-Ortiz, D.A.; Jaen-Cuellar, A.Y.; Delgado-Prieto, M.; Antonino-Daviu, J.A.
Power Disturbance Monitoring through Techniques for Novelty Detection on Wind Power and Photovoltaic Generation. *Sensors* **2023**, *23*, 2908.
https://doi.org/10.3390/s23062908

**AMA Style**

Gonzalez-Abreu AD, Osornio-Rios RA, Elvira-Ortiz DA, Jaen-Cuellar AY, Delgado-Prieto M, Antonino-Daviu JA.
Power Disturbance Monitoring through Techniques for Novelty Detection on Wind Power and Photovoltaic Generation. *Sensors*. 2023; 23(6):2908.
https://doi.org/10.3390/s23062908

**Chicago/Turabian Style**

Gonzalez-Abreu, Artvin Darien, Roque Alfredo Osornio-Rios, David Alejandro Elvira-Ortiz, Arturo Yosimar Jaen-Cuellar, Miguel Delgado-Prieto, and Jose Alfonso Antonino-Daviu.
2023. "Power Disturbance Monitoring through Techniques for Novelty Detection on Wind Power and Photovoltaic Generation" *Sensors* 23, no. 6: 2908.
https://doi.org/10.3390/s23062908