Maintenance optimisation is essential to achieve cost-efficiency, availability and reliability of supply in electric power systems. The process of maintenance optimisation requires information about the costs of preventive and corrective maintenance, as well as the costs of failures borne by both electricity suppliers and customers. To calculate expected costs, information is needed about equipment reliability characteristics and the way in which maintenance affects equipment reliability. The aim of this Ph.D. work has been to develop equipment reliability models taking the effect of maintenance into account.

The research has focussed on the interrelated areas of condition estimation, reliability modelling and maintenance modelling, which have been investigated in a number of case studies. In the area of condition estimation two methods to quantitatively estimate the condition of disconnector contacts have been developed, which utilise results from infrared thermography inspections and contact resistance measurements. The accuracy of these methods were investigated in two case studies. Reliability models have been developed and implemented for SF6 circuit-breakers, disconnector contacts and XLPE cables in three separate case studies. These models were formulated using both empirical and physical modelling approaches. To improve confidence in such models a Bayesian statistical method incorporating information from the equipment design process was also developed. This method was illustrated in a case study of SF6 circuit-breaker operating rods. Methods for quantifying the effect of maintenance on equipment condition and reliability have been investigated in case studies on disconnector contacts and SF6 circuit-breakers. The input required by these methods are condition measurements and historical failure and maintenance data, respectively.

This research has demonstrated that the effect of maintenance on power system equipment may be quantified using available data. However, realising the full potential of these methods requires the gathering and utilisation of failure and maintenance data as well as condition measurements to be improved.

Power system maintenance optimisation involves obtaining the minimum total costs, including preventive and corrective maintenance costs and the cost of failures for both supplier and customer. To calculate the cost of failure, information is needed about the equipment reliability characteristics. It is also necessary to know how maintenance affects component reliability. The aim of the work leading up to this thesis has been to develop reliability models that include the effect of maintenance.

Three case studies have been carried out for different types of power system components using three distinct methods. In the first study the reliability of the first generation XLPE cables was modelled with respect to failures caused by water treeing using load-strength modelling. The model was based on assumptions of the ageing process and the distribution system characteristics. This study showed that it is possible to and overvoltage and insulation characteristics that can be fitted to agree with failure statistics for water tree ageing in XLPE cables. The second case study included a study of all circuit breaker failures in the Swedish transmission grid during the period from 1 January 1999 to 30 June 2003. In a subsequent investigation a method to combine information from the design process with maintenance records and failure statistics was employed using Bayesian methods. The resulting reliability model is continuously updated as more failure and maintenance data

becomes available. This case study showed that it is possible to develop reliability models for components that have not yet failed by utilising information from the design process and right-censored observations from inspections. Finally, in the third case study a quantitative method for establishing the condition of disconnector contacts by the use of thermography was developed. Two sets of measurements on disconnector contacts in the Swedish transmission

grid were carried out to establish the accuracy of the method. By utilising the results from the measurements estimates of the statistical distributions of the error sources were produced.

The results from the case studies show that the lack of detailed, high-quality data remains a critical problem when modelling reliability of power system equipment, even when using methods that require a minimum of data.