#
Unsupervised Monitoring System for Predictive Maintenance of High Voltage Apparatus^{ †}

^{*}

^{†}

## Abstract

**:**

## 1. Introduction

#### Contributions

## 2. Background: An Empirical Model for Electrical Aging Assessment

## 3. Detecting Aging Phenomena via Anomaly Detection

**w**and

**x**. Vector

**w**is computed by the same algorithm that characterizes a PD pattern, but phase information is discarded in this case. Toward that end, first a PD pattern is extracted from ${\mathbf{v}}_{t}$; it is represented as a matrix, whose columns correspond to the power supply phases, and rows mark the maximum amplitudes of the discharges. Thus, each element in the matrix identifies an amplitude-phase pair; the contents of that cell give the occurrences of discharges at that amplitude-phase pair in the time window $\delta $. The resolution of the analog-to-digital converter (ADCs) actually determines the quantization levels in both the amplitude range and the phase range. Figure 3 illustrates this mechanism in a toy example, where a 2-bit ADC is employed. As a result, the PD pattern has four levels of amplitudes ($A1$, $A2$, $A3$, $A4$) and as many levels of phases ($P1$, $P2$, $P3$, $P4$). In the example, only one discharge with a maximum amplitude $A1$ occurred at phase $P2$, whereas two discharges with the same amplitude occurred at phase $P4$ phase. Figure 3 gives the associate vector

**w**, which includes three elements with value $A1$, since phase information is discarded. Likewise, vector

**w**includes two elements with value $A2$, five elements with value $A3$, and one element with value $A4$. The resulting size of

**w**, in general, will depend on the non-zero occurrences of the PD pattern. The choice of not using phase information is related to the fact that such information is useful when the goal is to recognize the category of defect that generated the PD. However, the proposed strategy only focuses on the detection of aging phenomena.

**x**holds the histogram of the amplitudes of raw data. In that histogram, each bin covers a range of amplitudes. Given the fullscale $FS$ adopted to sample the signal, the bins are obtained by uniform quantization of such interval. The number of bins $nbins$ and $FS$ are two input parameters.

**w**and

**x**. A pair of well-known statistical tests, namely Chi-Square (Chi2) and Kolmogorov–Smirnov (KS) support this task. Chi2 and KS have been selected as (1) they can support non-parametric tests (i.e., no assumptions on the specific probability distributions are involved) and (2) they allow a real-time implementation of hypothesis testing even on low-cost, low power and resource-constrained embedded devices. The null hypothesis is that the signal measured at time ${T}^{*}$ and any signal measured before ${T}^{*}$ come from the same population. In other words, an anomaly occurs when the measurement at time ${T}^{*}$ is not consistent with previous measurements. The underlying hypothesis is that aging phenomena lead progressively to significant changes in the distribution of PDs. Remarkably, such discontinuities can be detected even in the absence of trained classifiers or any knowledge base.

**x**and

**w**, respectively.

**x**(measured at time ${T}^{*}$) and $\tilde{\mathit{x}}$ (measured before time ${T}^{*}$). Then, let ${O}_{i}$ be the value of the histogram of $\mathit{x}$ for the ith bin, whereas ${\tilde{R}}_{i}$ is the value of the histogram of $\tilde{\mathit{x}}$ in the ith bin. Finally, the quantity ${\chi}^{2}$ is worked out as

## 4. Online Monitoring for Predictive Maintenance

## 5. IoT-Based Predictive System

## 6. Experimental Results

#### 6.1. Preliminary Experiments: Setting the Resolution

#### 6.2. Experimental Session

#### 6.3. Computational Cost

## 7. Conclusions

- the unsupervised system properly detects the changes in the status of the specimens, thus enabling a maintenance schedule that can avoid unwanted breakdowns.
- When adopting the Chi2 test to implement anomaly detection a systematic pattern can be identified. After a transient period corresponding to the first part of the specimen’s life, the alerts assumes a periodical trend, which ends when the aging starts to severely affect the specimen.
- The adoption of a coarser resolution in the ADC may somewhat increase the sensitivity of Chi2 test, which in turn may be adjusted by properly tuning the $nbins$ parameter.
- The monitoring systems has real-time performance on a low-resources device.

## Author Contributions

## Funding

## Conflicts of Interest

## Abbreviations

AC | Alternate Current |

ADC | Analog Digital Converter |

Chi2 | Chi–Square |

CNN | Convolutional Neural Network |

DC | Direct Current |

DNN | Deep Neural Network |

GIS | Gas Insulated System |

HFCT | High Frequency Current Transformer |

IoT | Internet of Things |

KS | Kolmogorov–Smirnov |

PD | Partial Discharge |

UHF | Ultra High Frequency |

## References

- De Faria, H., Jr.; Costa, J.G.S.; Olivas, J.L.M. A review of monitoring methods for predictive maintenance of electric power transformers based on dissolved gas analysis. Renew. Sustain. Energy Rev.
**2015**, 46, 201–209. [Google Scholar] [CrossRef] - Selcuk, S. Predictive maintenance, its implementation and latest trends. Proc. Inst. Mech. Eng. Part J. Eng. Manuf.
**2017**, 231, 1670–1679. [Google Scholar] [CrossRef] - Stone, G.C. A perspective on online partial discharge monitoring for assessment of the condition of rotating machine stator winding insulation. IEEE Electr. Insul. Mag.
**2012**, 28, 8–13. [Google Scholar] [CrossRef] - Luo, Y.; Li, Z.; Wang, H. A review of online partial discharge measurement of large generators. Energies
**2017**, 10, 1694. [Google Scholar] [CrossRef][Green Version] - High-Voltage Test Techniques—Partial Discharge Measurements; International Standard IEC 60270:2000; International Electrotechnical Commission: Geneva, Switzerland, 2000.
- Wu, M.; Cao, H.; Cao, J.; Nguyen, H.L.; Gomes, J.B.; Krishnaswamy, S.P. An overview of state-of-the-art partial discharge analysis techniques for condition monitoring. IEEE Electr. Insul. Mag.
**2015**, 31, 22–35. [Google Scholar] [CrossRef] - Fofana, I.; Hadjadj, Y. Electrical-based diagnostic techniques for assessing insulation condition in aged transformers. Energies
**2016**, 9, 679. [Google Scholar] [CrossRef] - Li, S.; Li, J. Condition monitoring and diagnosis of power equipment: Review and prospective. High Voltage
**2017**, 2, 82–91. [Google Scholar] [CrossRef] - Gao, W.; Ding, D.; Liu, W. Research on the typical partial discharge using the UHF detection method for GIS. IEEE Trans. Power Deliv.
**2011**, 26, 2621–2629. [Google Scholar] [CrossRef] - Wang, M.H. Partial discharge pattern recognition of current transformers using an ENN. IEEE Trans. Power Deliv.
**2005**, 20, 1984–1990. [Google Scholar] [CrossRef] - Hao, L.; Lewin, P.; Hunter, J.; Swaffield, D.; Contin, A.; Walton, C.; Michel, M. Discrimination of multiple PD sources using wavelet decomposition and principal component analysis. IEEE Trans. Dielectr. Electr. Insul.
**2011**, 18, 1702–1711. [Google Scholar] [CrossRef] - Ester, M.; Kriegel, H.P.; Sander, J.; Xu, X. A density-based algorithm for discovering clusters in large spatial databases with noise. Kdd
**1996**, 96, 226–231. [Google Scholar] - Álvarez, F.; Garnacho, F.; Ortego, J.; Sánchez-Urán, M. Application of HFCT and UHF sensors in on-line partial discharge measurements for insulation diagnosis of high voltage equipment. Sensors
**2015**, 15, 7360–7387. [Google Scholar] [CrossRef][Green Version] - Gianoglio, C.; Ragusa, E.; Bruzzone, A.; Gastaldo, P.; Torello, E.; Guastavino, F. Tensor Based Algorithm for Automatic Partial Discharges Pattern Classification. In Proceedings of the 2019 21st European Conference on Power Electronics and Applications (EPE’19 ECCE Europe), Genova, Italy, 3–5 September 2019; pp. P.1–P.8. [Google Scholar]
- Catterson, V.; Sheng, B. Deep neural networks for understanding and diagnosing partial discharge data. In Proceedings of the 2015 IEEE Electrical Insulation Conference (EIC), Seattle, WA, USA, 7–10 June 2015; pp. 218–221. [Google Scholar]
- Song, H.; Dai, J.; Sheng, G.; Jiang, X. GIS partial discharge pattern recognition via deep convolutional neural network under complex data source. IEEE Trans. Dielectr. Electr. Insul.
**2018**, 25, 678–685. [Google Scholar] [CrossRef] - Li, G.; Rong, M.; Wang, X.; Li, X.; Li, Y. Partial discharge patterns recognition with deep Convolutional Neural Networks. In Proceedings of the 2016 International Conference on Condition Monitoring and Diagnosis (CMD), Xi’an, China, 25–28 September 2016; pp. 324–327. [Google Scholar]
- Dey, D.; Chatterjee, B.; Dalai, S.; Munshi, S.; Chakravorti, S. A deep learning framework using convolution neural network for classification of impulse fault patterns in transformers with increased accuracy. IEEE Trans. Dielectr. Electr. Insul.
**2017**, 24, 3894–3897. [Google Scholar] [CrossRef] - Li, G.; Wang, X.; Li, X.; Yang, A.; Rong, M. Partial discharge recognition with a multi-resolution convolutional neural network. Sensors
**2018**, 18, 3512. [Google Scholar] [CrossRef][Green Version] - Wan, X.; Song, H.; Luo, L.; Li, Z.; Sheng, G.; Jiang, X. Pattern recognition of partial discharge image based on one-dimensional convolutional neural network. In Proceedings of the IEEE 2018 Condition Monitoring and Diagnosis (CMD), Perth, WA, Australia, 23–26 September 2018; pp. 1–4. [Google Scholar]
- Wang, Y.; Yan, J.; Sun, Q.; Li, J.; Yang, Z. A MobileNets Convolutional Neural Network for GIS Partial Discharge Pattern Recognition in the Ubiquitous Power Internet of Things Context: Optimization, Comparison, and Application. IEEE Access
**2019**, 7, 150226–150236. [Google Scholar] [CrossRef] - Banno, K.; Nakamura, Y.; Fujii, Y.; Takano, T. Partial Discharge Source Classification for Switchgears with Transient Earth Voltage Sensor Using Convolutional Neural Network. In Proceedings of the IEEE 2018 Condition Monitoring and Diagnosis (CMD), Perth, WA, Australia, 23–26 September 2018; pp. 1–5. [Google Scholar]
- Tenbohlen, S.; Coenen, S.; Djamali, M.; Müller, A.; Samimi, M.H.; Siegel, M. Diagnostic measurements for power transformers. Energies
**2016**, 9, 347. [Google Scholar] [CrossRef] - Yazici, B. Statistical pattern analysis of partial discharge measurements for quality assessment of insulation systems in high-voltage electrical machinery. IEEE Trans. Ind. Appl.
**2004**, 40, 1579–1594. [Google Scholar] [CrossRef] - Jannah, R.R.; Khayam, U. Design, implementation, and testing of partial discharge signal pattern recognition and judgment system application using statistical method. In Proceedings of the IEEE ICEVT & IMECE, Surakarta, Indonesia, 4–5 November 2015; pp. 314–318. [Google Scholar]
- Guastavino, F.; Dardano, A.; Torello, E. Measuring partial discharges under pulsed voltage conditions. IEEE Trans. Dielectr. Electr. Insul.
**2008**, 15, 1640–1648. [Google Scholar] [CrossRef] - Guastavino, F.; Gianoglio, C.; Torello, E.; Ferraris, M.; Gianelli, W. Electrical Aging Tests on Conventional and Nanofilled Impregnation Resins. In Proceedings of the 2018 IEEE Conference on Electrical Insulation and Dielectric Phenomena (CEIDP), Cancun, Mexico, 21–24 October 2018; pp. 156–158. [Google Scholar]
- Ahmed, Z.; Kluss, J.; Wallace, D. Partial discharge measurements and techniques for pattern recognition and life prediction of medium voltage XLPE cables. In Proceedings of the 2018 IEEE International Conference on High Voltage Engineering and Application (ICHVE), Perth, WA, Australia, 23–26 September 2018; pp. 1–4. [Google Scholar]
- Caprara, A.; Cavallini, A.; Garagnani, L.; Guo, J. A novel approach for continuous monitoring of Partial Discharge phenomena on Medium Voltage equipments. In Proceedings of the 2018 IEEE Electrical Insulation Conference (EIC), San Antonio, TX, USA, 17–20 June 2018; pp. 495–498. [Google Scholar]
- Cavallini, A.; Contin, A.; Montanari, G.C.; Puletti, F. Advanced PD inference in on-field measurements. I. Noise rejection. IEEE Trans. Dielectr. Electr. Insul.
**2003**, 10, 216–224. [Google Scholar] [CrossRef] - Cavallini, A.; Conti, M.; Contin, A.; Montanari, G. Advanced PD inference in on-field measurements. II. Identification of defects in solid insulation systems. IEEE Trans. Dielectr. Electr. Insul.
**2003**, 10, 528–538. [Google Scholar] [CrossRef] - IEEE Guide for Diagnostic Field Testing of Electric Power Apparatus—Electrical Machinery; International Standard IEEE 62.2:2004; IEEE Power and Energy Society: Piscataway, NJ, USA, 2004.
- Chandola, V.; Banerjee, A.; Kumar, V. Anomaly detection: A survey. ACM Comput. Surv. (CSUR)
**2009**, 41, 1–58. [Google Scholar] [CrossRef] - Aissa, N.B.; Guerroumi, M. Semi-supervised statistical approach for network anomaly detection. Procedia Comput. Sci.
**2016**, 83, 1090–1095. [Google Scholar] [CrossRef][Green Version] - Ye, N.; Chen, Q. An anomaly detection technique based on a chi-square statistic for detecting intrusions into information systems. Qual. Reliab. Eng. Int.
**2001**, 17, 105–112. [Google Scholar] [CrossRef] - Yu, M. A nonparametric adaptive CUSUM method and its application in network anomaly detection. Int. Adv. Comput. Technol.
**2012**, 4, 280–288. [Google Scholar] - Manikopoulos, C.; Papavassiliou, S. Network intrusion and fault detection: A statistical anomaly approach. IEEE Commun. Mag.
**2002**, 40, 76–82. [Google Scholar] [CrossRef] - Goldstein, M.; Dengel, A. Histogram-Based Outlier Score (Hbos): A Fast Unsupervised Anomaly Detection Algorithm; German Research Center for Artificial Intelligence (DFKI): Kaiserslautern, Germany, 2012; pp. 59–63. [Google Scholar]
- Massey, F.J., Jr. The Kolmogorov-Smirnov test for goodness of fit. J. Am. Stat. Assoc.
**1951**, 46, 68–78. [Google Scholar] [CrossRef] - Gianoglio, C.; Guastavino, F.; Ragusa, E.; Bruzzone, A.; Torello, E. Hardware Friendly Neural Network for the PD Classification. In Proceedings of the 2018 IEEE Conference on Electrical Insulation and Dielectric Phenomena (CEIDP), Cancun, Mexico, 21–24 October 2018; pp. 538–541. [Google Scholar]
- Ragusa, E.; Gianoglio, C.; Gastaldo, P.; Zunino, R. A digital implementation of extreme learning machines for resource-constrained devices. IEEE Trans Circuits Syst. II Express Briefs
**2018**, 65, 1104–1108. [Google Scholar] [CrossRef] - Ragusa, E.; Gianoglio, C.; Zunino, R.; Gastaldo, P. A Design Strategy for the Efficient Implementation of Random Basis Neural Networks on Resource-Constrained Devices. Neural Process. Lett.
**2019**, 1–19. [Google Scholar] [CrossRef] - Raspberry Pi Foundation. Raspberry Pi 3 Model B+. Available online: https://www.raspberrypi.org/products/raspberry-pi-3-model-b-plus/ (accessed on 28 February 2020).

**Figure 2.**Processing unit entitled to compute both vector of the amplitudes and vector of occurrences in the bins from the raw data.

**Figure 4.**Flowchart of the online alert unit entitled to output an alert if a significant change is detected.

**Figure 5.**Monitoring system for alert detection and classification, and cloud infrastructure collecting data from several monitoring system.

**Figure 8.**Specimen 1: (

**a**) alerts generated by the 8 bit system; (

**b**) normalized apparent charge computed on PD Pattern and (

**c**) linear regression residuals on normalized apparent charge.

**Figure 9.**Specimen 2-6: (

**a**) alerts generated by the 8 bit system and (

**b**) normalized apparent charge computed on PD Pattern and (

**c**) linear regression residuals on normalized apparent charge.

**Table 1.**Estimated execution times for the computing blocks involved in the proposed monitoring system.

Processing Unit [ms] | Chi2 [ms] | KS [ms] | |
---|---|---|---|

Average Time | 2.3 | 1.7 | 60.3 |

Maximum Time | 4.4 | 42.4 | 132.0 |

© 2020 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 (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Gianoglio, C.; Ragusa, E.; Bruzzone, A.; Gastaldo, P.; Zunino, R.; Guastavino, F.
Unsupervised Monitoring System for Predictive Maintenance of High Voltage Apparatus. *Energies* **2020**, *13*, 1109.
https://doi.org/10.3390/en13051109

**AMA Style**

Gianoglio C, Ragusa E, Bruzzone A, Gastaldo P, Zunino R, Guastavino F.
Unsupervised Monitoring System for Predictive Maintenance of High Voltage Apparatus. *Energies*. 2020; 13(5):1109.
https://doi.org/10.3390/en13051109

**Chicago/Turabian Style**

Gianoglio, Christian, Edoardo Ragusa, Andrea Bruzzone, Paolo Gastaldo, Rodolfo Zunino, and Francesco Guastavino.
2020. "Unsupervised Monitoring System for Predictive Maintenance of High Voltage Apparatus" *Energies* 13, no. 5: 1109.
https://doi.org/10.3390/en13051109