Towards a Data-Driven Approach to Ground-Fault Location
Mauro Caporuscio1, Pieternella Cijvat2, Hans Ottosson3
Core research areas: Computer Science (Data-driven technology), Electrical Engineering (Power Grids)
Problem definition and value
Power reliability is a very important property for any power grid. Reliability can be achieved by means of pre-emptive or fault location techniques. Pre-emptive techniques aim at avoiding outages by leveraging pathways and equipment redundancy, which requires large investments. Fault location techniques aim at decreasing the time for fault clearance.
European networks provide a high level of reliability with a low number of faults and a short fault-clearance time (between 15 to 400 min per customer/year) [1]. However, since about 80% of the faults observed in power networks occur at the distribution level, fault location in distribution networks has become an area of high interest for both academic and industrial research, previously reserved to transmission networks [1].
Ground faults are very challenging to locate. Traditionally, the fault location process is initiated by customers, which notify an operator about an outage [8]. The trouble call only indicates one specific location compromised by the outage, and additional calls are needed to approximate the affected area. This information is then combined with knowledge about the network and positions of fault clearing devices (e.g., Circuit Breaker) to identify possible fault locations. Relying on trouble calls as a fault indicator has a number of shortcomings: (1) customers tend to postpone the fault report, (2) reports are usually incomplete, (3) customers do mistakes and report about false outages, and (4) faults occurring during night-time are less likely to be reported. Once a fault is verified and confirmed, a technician is usually dispatched to check, classify, and recover the fault [2]. This can be a tedious and resource-consuming task as the area in which the fault occurred might be very large and the whole process usually depends on acquired knowledge about the area, such as previous experience and historical data [2][8].
To this extent, data analysis can help to improve the accuracy of the fault location, reduce the time needed to fix the problem, and optimize the resources [8]. Main methods for fault location through data analysis can be categorized as:
Impedance and Other Fundamental Frequency Component-Based Methods: These methods are widely used in distribution systems because of their cost-effectiveness. The methods require data about line impedance, voltage and current that are typically collected at the substation level, and make use of the fundamental frequency to approximate the fault location. The method computes multiple estimations by identifying a number of possible fault locations. In fact, the impedance is calculated starting from the measuring point and identifies all possible points in the network with an impedance equal to the assumed fault impedance.
High Frequency Components and Traveling Wave Based Methods: The view of this method was based on the reflection and transmission of the fault generated travelling waves on the faulted power network.
Although in this technique the fault can be located with high accuracy, the implementation is complex and more expensive than the implementation of impedance-based techniques.
Recently, some knowledge-based methods have been proposed to reduce the number of real-time calculations and, therefore, the computational load. In contrast to traditional techniques that strive for precision and correctness, knowledge-based approaches aim at finding complex correlations at the price of accuracy, i.e., resulting in a tradeoff between precision and uncertainty. There are three major families within the area of knowledge-based approaches: Artificial Neural Networks [3, 4], Fuzzy-logic Systems [5,6], and Support Vector Regression [7].
1 mauro.caporuscio@lnu.se - Linnaeus University (Dep. of Computer Science and Media Technology)
2 ellie.cijvat@lnu.se - Linnaeus University (Dep. of Physics and Electrical Engineering)
3 hans.ottosson@hughespowersystem.com - HUGHES Power System
Approach & Objectives
Departing from the aforementioned approaches that are not good enough for real commercial operation, we aim at developing a new method that combines Artificial Intelligence (AI) and Digital Twin (DT) technologies to locate or calculate the distance to fault position.
In the last recent years AI methods, especially, data-driven machine learning (ML) methods, have been applied to many fields of engineering and proved applicable in hard classification, regression and prediction problems. It is the aim of this SEED project to apply these methods to the Ground-fault location problem described above. The ML methods work best when trained with big data sets. This, however, is hard to get hands-on in the problem of fault localization, as faults and their locations haven’t been documented intensely in the past, and the data cannot be gained in controlled experiments (with reasonable costs). This is where DT technology comes into play: DTs allow for representing all physical elements (and their connections) of a given system in virtual (digital) models, which in turn is used to continuously simulate, observe and analyze the physical system’s behavior as a whole. Hence, we will make a DT of a distribution network and simulate its behavior in the presence of faults injected at known locations. Then, we observe the change of physical parameters (such as voltage and phase shift) over time (1 second) and connect these observations with the ground truth. This way, we can generate cheaply sufficiently big data sets to train the ML model. Only the most promising models will then be tested against reality.
The objective of this SEED project is twofold and includes:
O1: Conduct a feasibility study including (1) a State-of-the-art Analysis and a GAP Analysis about existing solutions for the ground-fault location problem (from both academic and industrial point of view), and (2) a simulation-based testbed evaluating the approach’s potential of success.
O2: According to the results obtained in O1.1 and O1.2, respectively: (1) identify and specify the industrial needs for the future, and (2) plan the development and commercialization of the envisioned approach.
Future externally funded project
The long-term objective is to develop and prototypically implements a new method that combines Artificial Intelligence and Digital Twin technologies to find a new solution to the challenging ground- fault location problem. To this end, we aim at applying for external funding to Vinnova’s Smarter Electronic Systems 2020 call4.
Expected Results
1. A detailed reasoning about existing solutions for the ground-fault location problem (O1.1) 2. A simulation study proving the feasibility of the envisioned approach (O1.2)
3. A Requirements specification for future Smart Electronics products (O2.1)
4. A Project proposal, including implementation and commercialization plans, to be submitted for external funding to Vinnova’s Smarter Electronic Systems 2020 call (O2.2)
Consortium
Members of the project have the required expertise in the project-relevant domains. In particular, research is being at Linnaeus University together with the Hughes Power System domain specialists.
Hughes Power System (HPS) is a Swedish manufacturer and specializes in research, development, manufacturing, marketing and sales of medium voltage outdoor and indoor switchgear products. HPS has more than 30 years of international experience in medium voltage electrical distribution systems, and offer customers integrated solutions with innovative and reliable products. As experts in smart electronics production, marketing and sales, HPS will be responsible for the requirements analysis (O2.1) and the
4https://www.smartareelektroniksystem.se/
Vinnova Smarter Electronic Systems 2020 project proposal (O2.2). HPS will contribute to the state-of-the- art and the gap analyses (O1.1). HPS representative in the project is:
– Hans Ottosson is Project Manager specialist in medium voltage product design and protection relays.
Department of Computer Science and Media Technology (DM) is an international knowledge center where research is carried in tight collaboration with regional, national and international partners, in academia and industry and is regularly attracting external funding in competitive research grants, e.g., Vinnova, The Swedish Research Council, The Knowledge Foundation, and H2020. As experts in Smarter Systems, with particular emphasis on data-driven technologies, and cyber-physical systems modeling and simulation, DM be responsible for the development of the ML technique for Ground-Fault location to be tested in the simulation study (O1.2). The DM team constitutes of three members:
– Mauro Caporuscio is Associate Professor (Docent) in Computer Science and provides expertise in Cyber-Physical Systems, Self-Adaptive Systems and Model-Driven Software Engineering.
– Mirko D’Angelo is a Ph.D. student in Computer Science and provides key expertise in Cyber-Physical Systems Modelling and Simulation, Self-Adaptive Systems and Machine Learning.
– Welf Löwe is Professor of Computer Science and provides expertise in Software analysis, and optimization, as well as in Data-Intensive systems, Machine Learning and Artificial Intelligence.
Department of Physics and Electrical Engineering (IFE) at Linnaeus University has attracted research funding from the Swedish Research Council (SSF), the Swedish Energy Agency (Energimyndigheten), the Swedish Power Grid (Svenska Kraftnät), and EU H2020, amongst others. The research group Waves, systems and signals has, or recently had, projects related to grid stability and hydropower, EM modeling of HVAC cables, and SiC MOSFETs for switch-mode applications. As domain experts (i.e., Power Grids and ground-fault location), IFE will be responsible for the state-of-the-art and the gap analyses (O1.1), and contribute to the simulation study (O1.2). Participating from IFE:
– Pieternella Cijvat, PhD, is a university lecturer in Electrical Engineering with expertise on SiC switch- mode measurement systems and smart grid.
The actors above are well versed in working with research projects in general and in collaborating together in applied research and development. The participants are forming a well-welded group of people that can create results in any given situation and specifically so in this one. DM and HPS have recently finalized a four-year project financed by KKS-Synergy. The objective of this Synergy was to investigate how to apply Model-Driven Engineering and Self-adaptation techniques to improve the management and delivery of stable sustainable electric energy. Whereas, the SEED project addresses a different problem (i.e., ground- fault location), and aims at exploiting different types of techniques (e.g., data-driven approach).
Activities & Time plan
Figure 1. Time plan
The work is divided into Activities (A1-A3) specific to the research objectives of this project spanning over a period of 6 months, starting from December 2019 (see Figure 1):
A1: Analysis (leader IFE, participant DM). This Activity is devoted to conduct the State-of-the-art Analysis, and the GAP Analysis about existing solutions for ground-fault location. Deliverable D1 Current best practices for ground-fault Location addresses O1.1 will be released at month 3.
A2: Simulation and Data Analysis (DM, IFE). This activity is devoted to (1) setup the Power Grid simulation environment to create a realistic “fault location” dataset, and (2) develop a preliminary Machine Learning model for ground-fault Location. D2 Trained ML addresses O1.2 and will be released at month 6.
Jan 2020 Feb 2020 Mar 2020 Apr 2020 MAy 2020 Jun 2020
A1 A2 A3
A3: Project Proposal (HPS, DM, IFE). This activity is devoted to the development of a project proposal to be submitted for external funding. D3 Project Proposal addresses O2 and will be released according to the Vinnova Smarter Electronic Systems 2020 call (expected in March 2020).
Budget
The overall budget of the project is approximately 220.000 SEK. We are applying for 100.000 SEK, since the remaining 120.000 SEK will be covered by DM and IFE (LNU) and HPS. Main costs are reported in the following table.
Participant Item SEED2020 Co-financing TOTAL[SEK]
LNU
Student Internship: to support the
development of the research project 40.000 (DM) 20.000 60.000
R&D time - (LNU) 50.000 50.000
HPS R&D time (125 hours, 800 SEK/h) 50.000 (HPS) 50.000 100.000
Travels and Meetings organization 10.000 - 10.000
Total 100.000 120.000 220.000
References
[1] Brahma, S.M. Fault location in power distribution system with penetration of distributed generation. IEEE Transactions on Power Delivery, VOL. 26, NO 3. 2011.
[2] Bahmanyar, A., Jamali, S., Estebsari, A., Bompard, E. A comparison framework for distribution system outage and fault location methods. Electric Power System Research. Elsevier. 2017.
[3] D. Thukaram, H.P. Khincha, H.P. Vijaynarasimha. Artificial neural network and support vector machine approach for locating faults in radial distribution systems. IEEE Trans. Power Deliv. 20. 2005.
[4] S.A.M. Javadian, A.M. Nasrabadi, M.R. Haghifam, J. Rezvantalab. Determining fault’s type and accurate location in distribution systems with DG using MLP neural networks. In proc. of International Conference on Clean Electrical Power, Capri. 2009.
[5] J.J. Mora, G. Carrillo, L. Perez. Fault location in power distribution systems using ANFIS nets and current patterns. In Proc. of Transmission & Distribution Conference and Exposition, Latin America, Caracas. 2006.
[6] F. Chunju, K.K. Li, W.L. Chan, Y. Weiyong, Z. Zhaoning. Application of wavelet fuzzy neural network in locating single line to ground fault (SLG) in distribution lines. Int. J. Electr. Power Energy Syst. 29. 2007.
[7] L. Ye, D. You, X. Yin, K. Wang, J. Wu, An improved fault-location method for distribution system using wavelets and support vector regression. Int. J. Electr. Power Energy Syst. 55. 2014.
[8] E. F. Ferreira, D. Barros. Faults Monitoring System in the Electric Power Grid of Medium Voltage. Procedia Computer Science, Volume 130. 2018
