DECLARATION 2- PUBLICATIONS
2.6 The role of AM processes on risk mitigation
2.6.1 AM technologies and techniques
There are many ways of perceiving that risks exist within power utility AM systems. This can be in terms of: occurrence of unscheduled events, removal of components from service for
corrective action, failure, type of strategy used and operation of an aging power infrastructure system. The current approaches to risk assessment have been driven by the imperative to keep power infrastructure assets for as long as possible, even as long as double their design lives. For example, it has been shown that almost one half of transformers currently in the power grid have exceeded their 30-year design life, but economic constraints are likely to compel the power utilities to keep them in service for up to 60 years [4]. In fact, it has also been reported that the asset replacement cycle can reach up to 100 years [64]. The tendency to utilize an aging infrastructure implies that the risk profile will be on the rising trend, hence the rising demand for new models for assessing the risk profile and for mitigating the risk.
However, current models for reliability and risk mitigation have traditionally been based on repairable failures without incorporating aged, non-repairable failures. A new method to incorporate aging failures in power system reliability evaluation was presented in [21]. The inclusion of aging failures in reliability evaluation can avoid underestimation of system risk and of the most definite misleading conclusion in system planning that can result from the underestimation [21]. Therefore, aging assets such as transformers should be prioritized during the risk modelling process. The traditional way of prioritizing a large number of aging transformers is by screening them using a simple ranking by age, but a more comprehensive method can be achieved by a risk assessment method which also incorporates contingency plans for ensuring that the spare parts are available [4]. In addition, the optimal number of spares should be determined. There was no probabilistic model for estimating the optimal number of transformer spares for a long time. A probabilistic model for determining the optimal number of transformer spares, based on the Poisson probability, was recently developed in [65]. The determination of the optimal number of spares and usage of transformer spares are important in ensuring that the system withstands occasional, catastrophic failures so that an acceptable level of system reliability is sustained [65].
AM technologies in electric power can be divided into hardware based and software based technologies. Hardware based technologies involve use of physical equipment for fault diagnosis, or condition monitoring or repair. Software based technologies comprise information technology (IT) that normally comes in the form of decision support tools. The electric power sector has seen a leap in progression of technological advancements over the last two decades. The most common conventional technologies employed in condition monitoring include [25]: resistance to ground (RTG) testing, surge comparison (surge) testing, high voltage (Hi-Pot) testing, motor current balance analysis (MCBA), and partial discharge monitoring (PDM), whereas newer technologies consist of motor circuit analysis (MCrA), motor current signature analysis (MCSA), motor power or electrical signature analysis (MPA), motor flux analysis (MFA), motor normalized temperature
analysis, and time domain reflectometry (TDR). Decision support based technologies include:
artificial intelligence, rule-based systems and inference engines, fuzzy logic, model-based approaches, neural networks, and data mining or automated rule-extraction. A detailed discussion of the above technologies and techniques can be viewed in Appendix D.
Since good experts are rare, even if a condition monitoring program is in operation, failures still occur, defeating the very purpose for which the investment in CBM was made. This has led to the application of artificial intelligence techniques (AIT) like expert systems, artificial neural networks, fuzzy logic; and later on, to distributed artificial intelligence which subsequently evolved into agent technology or multi-agent system (MAS) [66]. For example, the latest trends in AM demand that each power transformer be fitted with a special diagnostic apparatus, hence fuzzy logic presents a good option of ensuring non-intrusive diagnostic approach on the transformers. This is on the precept that a transformer’s life is mainly dependent on the life of solid insulation and the life-limit is determined by thermal degradation of the paper winding resulting in reduced Degree of Polymerization (DP); and generation of carbon monoxide (CO), carbon dioxide (CO2), and furans compounds. These are typical by-products of the degradation which fuzzy logic detects [67], [68].
An example of transformer condition monitoring health indicator is provided in Appendix E.
A fuzzy logic based transformer insulation paper deterioration estimation (FLDPE) is an example of advanced, novel fuzzy-based approach that uses inference rules to estimate insulation paper conditions, even where standard ANSI or IEEE methods could not, through three phases of insulation paper diagnosis, namely: tentative selection of CO2, CO; mechanical-fit process; and estimation and optimization of insulation paper status [67]. Expert systems are needed to interpret Dissolved Gas Analysis (DGA) results because the conventional DGA techniques face a difficulty in diagnosing slow developing and slight faults [69].
Within the power utilities, quantities of interest with respect to condition monitoring are so many that in a large power plant, the number of parameters measured may be too many to handle and artificial neural nets (ANNs) have the capability to effectively handle such data sizes. ANNs can be used in both estimation and classification mode to give an on-line indication of the power transformer condition with respect to the physical integrity of the windings [70]. The results from the ANNs analysis can thus be used in the fault root cause analysis and in the mitigation of failure risk [4].
Open System Architecture Condition-Based Maintenance (OSA-CBM) and (Machinery Information Management Open Systems Alliance (MIMOSA) initiated the development of Information and Computer Technology (ICT) protocols in condition monitoring and maintenance [66]. The IEEE contributed to these ICT protocols, especially in the development of condition
monitoring transducers and in fault root cause analysis [4], [25]. However, synergy of the latest and proposed technologies to the existing ICT platforms is lacking [66].
It is generally believed that IT based technologies like Supervisory Control and Data Acquisition (SCADA) and Geographic Information System (GIS) are crucial for real-time monitoring, tracking asset condition and restoration of the faults that result in supply interruption due to the radial topology of the distribution systems, and they form an integral part of Automatic Control System (ACS) and Feeder Automation System (FAS) [56]. Although many numerical or mathematical models have been developed to prioritize power distribution asset maintenance activities, there is a need for mathematical models which represent the effect of maintenance on reliability, to find the optimal strategy for the RCM where IT applications can play a central role in the data repository and acquisition.
Web and agent technologies are the latest developments in the AIT. Table 2-2 demonstrates their applications in the electric power industries.
There are three major challenges associated with the application of ICTs in the electric power infrastructure AM [66]:
1) First, the major setback to the advancement of ICT application to condition monitoring and maintenance has been the unsystematic application, as firms fail to adopt the already existing research efforts like OSA-CBM and MIMOSA.
2) Second, despite different architectures, methodologies and tools being proposed by researchers for the development of agent systems, the use of mobile devices (agents) is still low.
3) Third, AITs are still at infancy stage and their application has been sporadic, limited and inconsistent; where some progress has been made, the technologies have either been at proposal or experimental stage, or have been applied to an isolated practical case (a particular type of machine).
Table 2-2: Application of ICT in the power sector [66]
Baseline technology Application
Multi-agent system
Monitoring of gas turbine start up sequence Gas insulated sub-station (GIS) monitoring
Condition monitoring of a gas turbine during start-up
Off-line monitoring of power transformers using UHFoPD (ultra- high frequency of partial discharge)
Diagnostic and condition monitoring applications Post-fault disturbance diagnosis in power systems
Condition tele-monitoring and diagnosis in a power system On-line condition monitoring of transformers
Integration of software systems
Monitoring and diagnosis in supervisory systems Transmission and distribution systems
Integration of maintenance systems Multi-agent system,
Web technology
Maintenance management
Mobile agents, Web technology
Circuit breaker maintenance
Web technology Monitoring and diagnostic of a remote power system
Multi-agent systems, Mobile agents, Web technologies
On-site monitoring of distributed transformers in a power grid Remote analysis and reporting functions for data stored in sub- station databases
Remote control of distributed systems
Issues of ICT and decision support have a great bearing in power utility risk-based AM [70].
Improved decision support is very useful in tackling challenging decisions that utility asset managers face [64]. However, the electric power sector experiences significant changes in business environment that, in turn, make management decisions increasingly challenging. The challenges are as follows [64]:
1) An aging infrastructure that is wearing out on a 30-year cycle and being replaced on a 100-year cycle;
2) An electricity demand growth rate that, with the current cost models, threatens shareholder value;
3) Dramatically shorter planning horizons accompanied by increasingly constraining planning considerations such as environmental factors;
4) Expectations from the financial community, utility asset managers and financial managers for improved risk characterization and management; and
5) Increasing pressure from customers and regulators to maintain or even enhance service reliability and to control or reduce costs.
Transformer AM is generally considered to be one of the most important parts of power system equipment AM because they take most of the investment and are a major factor that affects reliability of the power system [68]. In addition, they are an integral part of power systems, and their reliability directly affects the reliability of the whole network because their outages can only be in one of the two forms: either forced by operation of automatic switching of protection systems due to external or internal causes or scheduled [71]. The management of transformers consist of three major activities, namely: the application of condition monitoring techniques in the operation, performing maintenance plans whilst investigating the less costly methods, and assessing the health and end of life of the transformer [68].
AM strategies in the power distribution sector can be described based on short term (operational), mid-term (asset maintenance) and long-term (strategic planning); whereas the main maintenance categories include CM, TBM, CBM and RCM [2], [56]. The CBM leads to high availability with moderate maintenance costs and is mainly applied in Extra High Voltage (EHV) and in High Voltage (HV) grids but the strategy is slowly being employed in Medium Voltage (MV) level as well. Furthermore, traditionally, statistical analyses are suitable for determining the remaining technical life in networks that have a large number of components like in the distribution systems [2].
Reasons for transformer outages are geographical dependent since environmental conditions, loading levels, and qualification of maintenance crews are geographical dependent factors [71].
This implies that physical asset risk profiles are also geographical dependent, hence when analyzing the risk profiles, it is worth noting that assets from different operating regions should be evaluated separately.
The condition monitoring techniques discussed so far are a means of determining probabilities of failures during fault and event tree analysis. However, the best way of determining the failure probability is by the application of statistical models to life data [4]. Hence, Section 2.6.2 examines how the life data can be applied in the power distribution risk AM.