# State Monitoring and Fault Diagnosis of HVDC System via KNN Algorithm with Knowledge Graph: A Practical China Power Grid Case

^{1}

^{2}

^{3}

^{*}

## Abstract

**:**

## 1. Introduction

## 2. HVDC System Knowledge Graph

#### 2.1. Knowledge Acquisition

#### 2.2. Knowledge Analysis

#### 2.3. Knowledge Base Establishment

#### 2.4. Graph Construction

#### 2.5. Knowledge Service

#### 2.6. Knowledge Application

## 3. Fault Classification

#### 3.1. AC Fault

#### 3.2. DC Fault

#### 3.3. Converter Valve Fault

#### 3.4. Commutation Failure

## 4. Principle of KNN Algorithm

- (1).
- Mahalanobis distance

- (2).
- Chebyshev distance

- (3).
- ED

- (4).
- Manhattan distance

## 5. Fault Diagnosis Model

- (1)
- Data processing, normalize the data of 15 channels in each type of fault data as follows:$${x}^{*}=\frac{{x}_{i}-{x}_{\mathrm{min}}}{{x}_{\mathrm{max}}-{x}_{\mathrm{min}}}$$
- (2)
- The data of 15 channels of each sample are connected in series head to tail and stacked according to the number of samples of fault type to form all fault datasets;
- (3)
- Label the fault data;
- (4)
- Data classification: randomly divide 80% of all fault data into training sets and the remaining 20% into test sets;
- (5)
- Establish KNN fault diagnosis model, set the appropriate KNN algorithm parameter K value, and select the appropriate distance function;
- (6)
- 80% of data are substituted into the fault diagnosis model for fault diagnosis training, and the remaining 20% of data are substituted into the trained model for verification;
- (7)
- Obtain the test data label and compare the diagnosed label with the real label of the test data, and then calculate the fault diagnosis accuracy rate and draw a visual confusion matrix diagram. The accuracy rate formula is as follows.$$p=\frac{{N}_{\mathrm{label}-\mathrm{ture}}}{{N}_{\mathrm{label}-\mathrm{all}}}$$

## 6. Case Study

_{1}(n

_{1}= 2, n

_{2}= 3, n

_{3}= 3, n

_{4}= 4) of all fault datasets. The second group of test data are the training data themselves Y

_{2}(n

_{1}= 8, n

_{2}= 11, n

_{3}= 11, n

_{4}= 14). After training the model, the training data are substituted into the model for verification. The third group of test data are all fault data Y

_{3}(n

_{1}= 10, n

_{2}= 14, n

_{3}= 14, n

_{4}= 18).

_{1}test set is shown in Figure 10. The KNN algorithm has the highest classification accuracy among the three methods, with a classification accuracy of 83.3%. The diagnosis accuracy of the other two algorithms is lower than 80%; the KNN has two samples of diagnostic errors, the SVM algorithm has three samples of diagnostic errors, the BC algorithm has four samples of diagnostic errors, and the DC fault diagnosis accuracy of the three algorithms is relatively low.

_{2}test set is shown in Figure 11. The fault diagnosis accuracy of the KNN algorithm is 100%, and both the naive BC and SVM algorithms generate diagnostic errors; the fault diagnosis accuracy of SVM is 88.6%, and that of the BC algorithm is 75%. The diagnostic accuracy of the SVM algorithm is higher than that of the BC algorithm, and these two algorithms generate this error for DC fault.

_{3}test set are shown in Figure 12. It is obvious that the diagnosis results accuracy of the naive BC and SVM algorithms is not as high as that of the KNN algorithm; the fault accuracy rate of the KNN algorithm is 100%. Although the accuracy rate of the BC and SVM algorithms is lower than that of the KNN algorithm, the accuracy rate of these two algorithms is also higher than 80%. Moreover, from the final results of the three sets of test sets, it can be seen that, among the kinds of faults, DC faults readily cause classification errors.

## 7. Discussion and Limitations

#### 7.1. Discussion

_{1}. The training data themselves are used as test set Y

_{2}, and then all data are used as test set Y

_{3}. The purpose of this is to test all the data and achieve cross-validation. In general, it is most reasonable that the test dataset and training set are different.

#### 7.2. Limitations

## 8. Conclusions

## Author Contributions

## Funding

## Conflicts of Interest

## Nomenclature

AC | alternating current |

BC | Bayesian classifier |

CNN | convolutional neural networks |

DC | alternating current |

ED | Euclidean distance |

ES | expert system |

HVDC | High-voltage direct current |

KG | knowledge graph |

KNN | K-Nearest Neighbor |

SVM | support vector machine |

K | number of adjacent points |

d | the distance between samples |

N | the number of samples |

S_{i} | the standard deviation of the sample data in the i-th dimension |

x | the horizontal coordinate position of the sample |

y | the vertical coordinate position of the sample |

w | classification weight of samples |

x* | normalized sample data |

p | accuracy of fault diagnosis |

## References

- Mohamed, N.; Ahmed, E.; Tamou, N. Improving Low-Voltage Ride-Through Capability of a Multimegawatt DFIG Based Wind Turbine under Grid Faults. Prot. Control. Mod. Power Syst.
**2020**, 5, 370–382. [Google Scholar] - Yang, B.; Liu, B.; Zhou, H.; Wang, J.; Yao, W.; Wu, S.; Shu, H.; Ren, Y. A Critical Survey of Technologies of Large Offshore Wind Farm Integration: Summarization, Advances, and Perspectives. Prot. Control. Mod. Power Syst.
**2022**, 7, 1–32. [Google Scholar] [CrossRef] - Yang, B.; Li, Y.; Li, J.; Shu, H.; Zhao, X.; Ren, Y.; Li, Q. Comprehensive Summarization of Solid Oxide Fuel Cell: Control: A State-of-the-Art Review. Prot. Control. Mod. Power Syst.
**2022**, 7, 1–31. [Google Scholar] [CrossRef] - Chen, Y.; Yang, B.; Guo, Z.; Wang, J.; Zhu, M.; Li, Z.; Li, Q.; Yu, T. Dynamic Reconfiguration for TEG Systems under Heterogeneous Temperature Distribution via Adaptive Coordinated Seeker. Prot. Control. Mod. Power Syst.
**2022**, 7, 1–19. [Google Scholar] [CrossRef] - Tanmay, D.; Ranjit, R.; Kamal, K. Impact of the Penetration of Distributed Generation on Optimal Reactive Power Dispatch. Prot. Control. Mod. Power Syst.
**2020**, 5, 332–357. [Google Scholar] - Mehdi, T.; Mehdi, N. Human Reliability Analysis in Maintenance Team of Power Transmission System Protection. Prot. Control. Mod. Power Syst.
**2020**, 5, 270–282. [Google Scholar] - Liu, H.; Li, B.; Wen, W.; He, J.; Li, H.; Li, B.; Han, W. Review and Prospect on Transmission Line Protection in Flexible DC System. Power Syst. Technol.
**2021**, 45, 3463–3477. [Google Scholar] - Yang, B.; Sang, Y.; Shi, K.; Yao, W.; Jiang, L.; Yu, T. Design and Real-time Implementation of Perturbation Observer based Sliding-Mode Control for VSC-HVDC Systems. Control. Eng. Pract.
**2016**, 56, 13–26. [Google Scholar] [CrossRef] [Green Version] - Yang, B.; Jiang, L.; Yao, W.; Wu, Q. Perturbation Observer based Adaptive Passive Control for Damping Improvement of Multi-terminal Voltage Source Converter-based High Voltage Direct Current Systems. Trans. Inst. Meas. Control.
**2017**, 39, 1409–1420. [Google Scholar] [CrossRef] [Green Version] - Yang, B.; Yu, T.; Zhang, X.; Huang, L.; Shu, H.; Jiang, L. Interactive Teaching-learning Optimizer for Parameter Tuning of VSC-HVDC Systems with Offshore Wind Farm Integration. IET Gener. Transm. Distrib.
**2018**, 12, 678–687. [Google Scholar] [CrossRef] [Green Version] - Yang, B.; Jiang, L.; Yu, T.; Shu, H.; Zhang, C.; Yao, W.; Wu, Q. Passive Control Design for Multi-terminal VSC-HVDC Systems via Energy Shaping. Int. J. Electr. Power Energy Syst.
**2018**, 98, 496–508. [Google Scholar] [CrossRef] - Yan, B.; Tian, Z.; Shi, S.; Weng, Z. New Fault Diagnosis Method for HVDC Transmission System. Autom. Electr. Power Syst.
**2007**, 31, 57–61. [Google Scholar] - Sun, T. Fault Analysis and Line Protection Scheme of HVDC Transmission System. Electr. Power Technol.
**2015**, 4, 119–120. [Google Scholar] - Li, Z. Research on a New Method of HVDC System Fault Diagnosis Based on Auto Disturbance Rejection Controller. Telecom Power Technol.
**2010**, 27, 39–43. [Google Scholar] - Li, Z. Discussion on Protection Technology of DC Transmission System. Public Electr.
**2022**, 37, 43–44. [Google Scholar] - Xu, H. Fault Analysis and Protection of DC Transmission System. Technol. Innov. Appl.
**2017**, 17, 167–168. [Google Scholar] - Li, J.; Li, X.; Gao, T.; Zhang, J.; Zhang, B. Research and Application of Fault Handling Based on Power Grid Multivariate Information Knowledge Graph. Power Inf. Commun. Technol.
**2021**, 19, 30–38. [Google Scholar] - Zhang, R.; Liu, J.; Zhang, B.; Li, J.; Gao, T.; Zhang, J. Research on Grid Fault Handling Knowledge Graph Construction and Real-Time Auxiliary Decision Based on Transfer Learning. Power Inf. Commun. Technol.
**2022**, 20, 24–34. [Google Scholar] - Liu, F.; Shu, H.; Yang, D.; Yang, Y. Research on Fault Diagnosis Method of HVDC Transmission System. Yunnan Electr. Power
**2010**, 38, 23–25. [Google Scholar] - Shao, X.; Ning, Y.; Liu, Y.; Zhang, H. Review and Prospects of Fault Diagnosis in Power System. Ind. Control. Comput.
**2012**, 25, 4–7. [Google Scholar] - Yu, F.; Li, S.; Li, D.; Liu, X. Application of Support Vector Machine in Fault Diagnosis of HVDC Transmission System. Chin. J. Sci. Instrum.
**2006**, 27, 1734–1736. [Google Scholar] - Yu, F.; Li, S.; Ma, X.; Liu, X. Research on Fault Diagnosis Model of HVDC System Based on SVM. J. Syst. Simul.
**2007**, 19, 3180–3183. [Google Scholar] - Zhang, D.; Zhang, X.; Zhang, H.; He, J. Fault Diagnosis of AC/DC Transmission System Based on Convolutional Neural Network. Autom. Electr. Power Syst.
**2022**, 46, 132–140. [Google Scholar] - Liu, H.; Xing, C.; Chen, S.; Yang, H.; Yan, Z. High Voltage Direct Current Transmission Based on Elman Neural Network System Commutation Failure Fault Diagnosis. Softw. Guide
**2020**, 19, 9–14. [Google Scholar] - Zheng, X.; Peng, P. Fault Diagnosis of Flexible HVDC Converter Based on Preferred Wavelet Packet and AdaBoost-SVM. J. Power Syst. Autom.
**2019**, 31, 42–49. [Google Scholar] - Chen, C.; Chen, S.; Bi, G.; Gao, J.; Zhao, X.; Li, L. Fault Diagnosis of Weak Receiving DC Transmission System Based on Parallel CNN-LSTM. Mot. Control. Appl.
**2022**, 49, 83–91. [Google Scholar] - Huang, H. Application of KNN Algorithm in Single Terminal Fault Detection of HVDC Transmission. Energy Dev.
**2021**, 2, 22–27. [Google Scholar] - Lu, D.; Ning, Q.; Yang, X. Fault Diagnosis of Rolling Bearing Based on KNN-Naive Bayes Algorithm. Comput. Meas. Control.
**2018**, 26, 21–27. [Google Scholar] - Torres, O.; Alejandro, G. Grid Integration of Offshore Wind Farms Using a Hybrid HVDC Composed by an MMC with an LCC-Based Transmission System. Energy Procedia
**2017**, 137, 391–400. [Google Scholar] [CrossRef] - Shu, N.; Ge, Z.; Luo, J.; Chen, K.; Ding, S. Fault Diagnosis System Based on Knowledge Graph. Comput. Sci. Technol.
**2021**, 39, 11–13. [Google Scholar] - Liu, X.; Song, C.; Sun, L.; Ma, F. Research on Fault Intelligent Diagnosis Method Based on Knowledge Graph. Shandong Commun. Technol.
**2019**, 39, 18–20. [Google Scholar] - Liu, R.; Xie, G.; Yuan, Z.; Song, W.; Wang, G. Research on Intelligent Fault Diagnosis Based on Knowledge Graph. Posts Telecommun. Des. Technol.
**2020**, 10, 30–35. [Google Scholar] - Zheng, R.; Hu, Z.; Wen, Z.; Wang, J. AC Fault Detection Method for HVDC System. Guangdong Electr. Power
**2020**, 33, 97–104. [Google Scholar] - Li, J.; Qian, J.; Li, J.; Shan, J. Influence of AC Side Fault of Converter Station on HVDC System. Yunnan Electr. Power
**2007**, 35, 10–12. [Google Scholar] - Fan, S.; Yang, H.; Xiang, X.; Yang, H.; Li, W.; He, X. Modular Multilevel with DC Fault Ride through Capability Conversion Topology Deduction and Comparison. Electr. Power
**2021**, 54, 38–45. [Google Scholar] - Song, X.; Gao, Y.; Ding, G.; Yan, H. Lightning Strike Interference and Fault Identification of Transmission System. Insul. Surge Arresters
**2021**, 1, 96–110. [Google Scholar] - Narendra, K.; Sood, V.; Khorasani, K.; Patel, R. Application of a Radial Basis Function (RBF) Neural Network for Fault Diagnosis in a HVDC System. IEEE Trans. Power Syst.
**1998**, 13, 177–183. [Google Scholar] [CrossRef] - Liu, X.; Dai, D.; Rao, H.; Cheng, C.; Ai, L.; Wu, S. Common Faults and Treatment Methods of Converter Valve in DC System. Electron. World
**2015**, 21, 115–117. [Google Scholar] - Yuan, Y.; Wei, Z.; Lei, X.; Wang, W.; Sun, G. Survey of Commutation Failures in DC Transmission Systems. Electr. Power Autom. Equip.
**2013**, 33, 140–147. [Google Scholar] - Wei, J.; Lin, L.; Cheng, T.; Mou, D. Commutation Failure Factors Analysis in HVDC Transmission System. J. Chongqing Univ.
**2006**, 29, 16–18. [Google Scholar] - Chen, S.; Hua, X.; Xiang, W. Fault Diagnosis Method of Turbine Flow Passage Based on Improved KNN Algorithm and Its Application. Therm. Power Gener.
**2021**, 50, 84–90. [Google Scholar] - Bhattacharya, G.; Ghosh, K.; Chowdhury, A. An Affinity-Based New Local Distance Function and Similarity Measure for KNN Algorithm. Pattern Recognit. Lett.
**2012**, 33, 356–363. [Google Scholar] [CrossRef] - Maesschalck, R.; Rimbaud, D.; Massart, D. The Mahalanobis Distance. Chemom. Intell. Lab. Syst.
**2000**, 50, 1–18. [Google Scholar] [CrossRef] - Maurer, C.; Qi, R.; Raghavan, V. A Linear Time Algorithm for Computing Exact Euclidean Distance Transforms of Binary Images in Arbitrary Dimensions. IEEE Trans. Pattern Anal. Mach. Intell.
**2003**, 25, 265–270. [Google Scholar] [CrossRef] - Wang, Z.; Xu, K.; Hou, Y. Classification of Iris Based on KNN Algorithm with Different Distance Formulas. Wirel. Internet Technol.
**2021**, 13, 105–106. [Google Scholar] - Available online: http://prpq.nea.gov.cn/gzdt/4638.html (accessed on 18 September 2020).
- Kalra, P. Fault Diagnosis for an HVDC System: A Feasibility Study of an Expert System Application. Electr. Power Syst. Res.
**1988**, 14, 83–89. [Google Scholar] [CrossRef] - Liu, H.; Liu, J. Fault Diagnosis of Power Transformer Based on Graph Convolutional Network. J. Hunan Univ. Sci. Technol.
**2021**, 36, 75–81. [Google Scholar] - Mejdoub, M.; Amar, C. Classification Improvement of Local Feature Vectors over The KNN Algorithm. Multimed. Tools Appl.
**2013**, 64, 197–218. [Google Scholar] [CrossRef]

**Figure 8.**Waveforms of HVDC system four types of faults. (

**a**) AC fault data; (

**b**) DC fault data; (

**c**) converter valve fault data; (

**d**) commutation failure fault data.

**Figure 10.**Confusion matrix of experimental results for test set Y

_{1.}(

**a**) KNN model test result diagram; (

**b**) SVM model test result diagram; (

**c**) BC model test result diagram.

**Figure 11.**Confusion matrix of experimental results for test set Y

_{2.}(

**a**) KNN model test result diagram; (

**b**) SVM model test result diagram; (

**c**) BC model test result diagram.

**Figure 12.**Confusion matrix of experimental results for test set Y

_{3.}(

**a**) KNN model test result diagram; (

**b**) SVM model test result diagram; (

**c**) BC model test result diagram.

1: Establish KNN algorithm model; |

2: Set KNN algorithm parameters: K, $d$; 3: Import data and select the test set; 4: Calculate similarity; 5: Calculate the distance between training data and unknown data according to the selected $d$; 6: Calculate weight and judge similarity according to $w\left({x}_{i},{y}_{i}\right)$; 7: select front K data; 8: Record the times of each category; 9: Use the category with the most occurrences as the category of unknown data; 10: Repeated to judge all test data. |

Signal | Description Meaning | Signal | Description Meaning |
---|---|---|---|

UACA(V) | A-phase AC voltage | IACD_L_{3}(A) | C-phase AC current of D-bridge valve side |

UACB(V) | B-phase AC voltage | UDL(V) | DC line voltage |

UACC(V) | C-phase AC voltage | UDN(V) | Neutral bus voltage |

IACY_L_{1}(A) | A-phase AC current of Y-bridge valve side | IDN(A) | Neutral bus current |

IACY_L_{2}(A) | B-phase AC current of Y-bridge valve side | IDE(A) | Grounding pole bus current |

IACY_L_{3}(A) | C-phase AC current of Y-bridge valve side | IDH(A) | High-voltage bus current |

IACD_L_{1}(A) | A-phase AC current of D-bridge valve side | IDL(A) | DC line current |

IACD_L_{2}(A) | B-phase AC current of D-bridge valve side |

Method | Parameter Name | Parameter Setting |
---|---|---|

KNN | Neighbors: K | 7 |

Metric distance | Euclidean distance | |

Weight type | Inverse distance | |

SVM | Penalty coefficient: C | 1 |

Kernel | Gaussian | |

Decision function shape | One-versus-one | |

BC | Nuclear type | Gaussian |

Test Sample | Number of Samples | Number of Positive Samples | Number of Negative Data | Accuracy | ||||||
---|---|---|---|---|---|---|---|---|---|---|

KNN | SVM | BC | KNN | SVM | BC | KNN | SVM | BC | ||

Y_{1} | 12 | 10 | 9 | 8 | 2 | 3 | 4 | 83.3% | 75% | 66.7% |

Y_{2} | 44 | 44 | 39 | 33 | 0 | 5 | 11 | 100% | 88.6% | 75% |

Y_{3} | 56 | 56 | 48 | 50 | 0 | 8 | 6 | 100% | 85.7% | 89.3% |

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**

Chen, Q.; Li, Q.; Wu, J.; He, J.; Mao, C.; Li, Z.; Yang, B.
State Monitoring and Fault Diagnosis of HVDC System via KNN Algorithm with Knowledge Graph: A Practical China Power Grid Case. *Sustainability* **2023**, *15*, 3717.
https://doi.org/10.3390/su15043717

**AMA Style**

Chen Q, Li Q, Wu J, He J, Mao C, Li Z, Yang B.
State Monitoring and Fault Diagnosis of HVDC System via KNN Algorithm with Knowledge Graph: A Practical China Power Grid Case. *Sustainability*. 2023; 15(4):3717.
https://doi.org/10.3390/su15043717

**Chicago/Turabian Style**

Chen, Qian, Qiang Li, Jiyang Wu, Jingsong He, Chizu Mao, Ziyou Li, and Bo Yang.
2023. "State Monitoring and Fault Diagnosis of HVDC System via KNN Algorithm with Knowledge Graph: A Practical China Power Grid Case" *Sustainability* 15, no. 4: 3717.
https://doi.org/10.3390/su15043717