Qikai Zhuang

Life prediction of a full-scale transformer winding insulation through statistical analysis of AC voltage endurance test data

It is now acknowledged that for many solid dielectrics below a threshold electrical stress, electrical aging can either be neglected or considered to progress very slowly. Hence, it is recommended to exclude from life prediction data obtained at high electrical fields. In general, it is common to be faced with an insufficient number of life data obtained at low fields. Therefore, we first introduce the Inverse Power Weibull model, which can estimate statistically the confidence bounds on the lifetime at a given reliability at operating voltage. Then, as the main contribution of this paper, the existing models pertaining to the size enlargement effect are extended to a non-uniform stress situation, and two methods are developed to calculate the confidence bounds of the lifetime of a full-scale high voltage component. Finally, the Inverse Power Weibull model and the calculation method we derived are applied to the life data collected from transformer winding insulation specimens. The voltage versus lifetime characteristics of the epoxy resin insulation of the windings comply with earlier research. After completion of the statistical analysis, a warranty life is suggested for the winding of a full-size transformer, with a high reliability and confidence level.

Aging of oil-impregnated transformer insulation studied through partial discharge analysis

The aging of the epoxy-paper insulation of an oil-impregnated transformer was studied through partial discharge analysis. Accelerated electrical aging tests were carried out on samples of transformers windings, with and without paper insulation. Partial discharge (PD) measurements according to standard IEC 60270 were performed periodically during aging, and partial discharge phase patterns were recorded throughout the aging test. Some representative PD phenomena were obtained from the experimental results. The evolution of typical PD patterns is reported until breakdown.

Statistical approach to establish failure behaviour on incomplete asset lifetime data

Asset failures, that needs to be managed, has an uncertain characteristic and analysis of uncertainty is essential to Asset Management (AM). Forecasting the technical performance of assets forms an integral part of strategic and operational activities within AM. To establish the failure behaviour of assets requires a significant degree of reliable asset information, which, in many practical cases, is not sufficiently rich or available to provide a basis for straightforward decision-making. In this paper a practical and systematic statistical methodology is used for dealing with incomplete asset lifetime data. The method described in this paper is based on a statistical parametric method and is applied with the aim of obtaining an indicator of the future failure expectancy with a certain confidence interval. On the whole, the paper concludes that, even though input data was either missing or incomplete, it is in certain cases possible to develop sensible probability models. These models take into account uncertainty and ultimately can be applied to facilitate the asset manager in AM decision-making. In addition to applying statistical methods, this contribution highlights the vital role of engineering and expert knowledge in interpreting the statistical results.

Life prediction for epoxy resin insulated transformer windings through accelerated aging tests

This investigation aims to perform a probabilistic prediction of the lifetime of epoxy resin insulation of transformer windings. To support the prediction, several accelerated electrical aging tests were conducted at different voltage levels on epoxy resin insulated transformer winding specimens. The lifetimes achieved from the tests show a “flat z-shaped” V-t (voltage vs. lifetime) curve and indicate the well known threshold stress effect. Since the operation stress is in the convex part of the V-t curve, the life exponent n was not predetermined, but estimated from a data pool obtained at the lowest voltage levels. As a result, the operation life and its confidence intervals were predicted from the breakdown data at these low voltage levels.

Upcoming Role of Condition Monitoring in Risk-Based Asset Management for the Power Sector

The electrical power sector is stimulated to evolve under the pressures of the energy transition, the deregulation of electricity markets and the introduction of intelligent grids. In general, engineers believe that technologies such as monitoring, control and diagnostic devices, can realize this evolvement smoothly. Unfortunately, the contributions of these emerging technologies to business strategies remain difficult to quantify in straightforward metrics. Consequently, decisions to invest on these technologies are still taken in an ad hoc manner. This is far from the risk-based approach commonly recommended for asset management (AM). The paper introduces risk-based management as a guiding principle for maintenance management. Then, the triple-level AM model (strategic, tactical and operational) as the foundation to define risk-based AM is described. Afterwards, two categories of risks, one triggered by technical stimuli and the other by non-technical stimuli are introduced. It is shown that the main challenge of managing risks with technical stimuli is to have the ability to understand the technical cause of failures, which is located at the operational level within the triple-level AM model. One method to quantitatively understand the technical cause of failures is by means of condition diagnostic and monitoring technologies. Therefore, the aim of this paper is to clarify the potential contribution of condition diagnostic and monitoring technologies to risk-based decision making for the power sector. This paper shows that, in practice, the implementation of condition diagnostic and monitoring technologies is mainly driven by purely technical asset based considerations without evaluating the contribution to, for instance, risks. This paper provides a list of aspects in which condition diagnostic and monitoring may contribute to risk evaluation with technical stimuli. The listed aspects (which are: (1) asset specific condition data, (2) timely condition data and (3) predictive condition data) can be regarded as input for the probability of failure and as influencing input for the consequence of failure, hence benefiting quantitative risk studies and AM activities (such as condition assessment/maintenance or replacement). Finally, these benefits can be evaluated afterwards in a risk-based AM planning stage, so that asset managers can justify investments on necessary technical improvements of condition monitoring systems.

Life prediction for transformer winding insulated with epoxy resin and thickness-reduced paper through voltage endurance tests

This paper investigates the epoxy resin coated paper-oil insulation between windings wires of a transmission power transformer. Voltage endurance test has been performed separately on the epoxy-resin layer and on the complete insulation. The test result reveals the voltage vs. lifetime characteristics of the epoxy resin layer, which further supports the life prediction on the complete insulation. The predicted lifetime ensures the reliability of the investigated insulation design and further suggests the possibility to reduce the thickness of the paper insulation layer.

Developing the physical layer of future power grids: Required for automated asset management and renewable energy integration

Increased demand for power requires a grid that is equipped to facilitate distributed or large-distance bulk generation. Complying with long term goals, this future grid should be reliable, carbon neutral and sustainable. The main drivers for the upcoming changes in the power grids are the increase of renewable energy production and the aging of the high voltage electrical components. In order to determine the health state of these components, intelligence embedded in the components is required. This includes sensors, communication technology and associated models for interpretation and decision making. To realize an intelligent future grid, it is required that, besides high level smart grids concepts, an actual lower level physical layer is developed. Concrete suggestions will be discussed to implement market ready solutions for predictive health management which can cope with the changing environment of the future grid. Using these suggestions in a case, it is shown that the loss of remaining lifetime can be decreased during a standard 24 hour loading pattern.

Next generation polymeric HVDC cables – a quantum leap needed?

The attention given to high voltage direct current (HVDC) cables appears to be at an all-time high. In the July/August 2017 issue of IEEE Electrical Insulation Magazine, the focus was on different aspects of the design and testing of HVDC cable systems. There are several reasons for this, one of the most important probably being the worldwide attention on harvesting renewable energy at a scale never seen before. The prices of building large offshore windfarms in Europe as reflected in a recent normalized cost of electricity have unexpectedly dropped in the last year, as illustrated in Figure 1, which will further accelerate the development of HVDC cables to link offshore and onshore substations. The winning bid in 2016 for the Borssele II, 700-MW windfarm in the Netherlands was as low as 54.5€ (

A statistical aging model for paper oil insulation under repetitive transients

61)/MWh [1].

Use case modelling for risk based maintenance management system

AC voltage has been applied on oil impregnated paper with superimposed transients of repetition frequency ranging from 1 kHz to 10 kHz. The time-to-breakdown has been recorded. Then the “accelerated life test analysis” (ALTA) approach is applied to model the lifetime vs. repetition frequency relationship with inverse power, Eyring or exponential laws. During the “accelerated life test analysis”, the recorded life data at multiple repetition frequencies can be inputted simultaneously into the statistical aging model, and the parameters of each life vs. frequency relationship can be estimated. The estimation shows that the inverse power, Eyring or exponential relationships can be applied in our test frequency ranges, and extrapolated to higher or lower frequencies.