^{1,a}, L.A.N. Costa

^{1}, A.M. Teodoro Filho

^{1}, G.M.O. Salles

^{2,b}and A. Pinto

^{2}

^{1}Foundation for Development of Bauru/ Bauru, SP, Brazil.

^{a}gabriel.costa.lima@aremas.com.br

^{2}Copel Power Generation Company/Curitiba, Brazil.

^{b}gisele.salles@copel.com

The transmission of electricity involves large number of assets (transformers, cables, connectors, etc) over a large geographic area and demand a large capital investment. These assets are subject different environmental conditions such temperature, radiation, wind speed, etc. As result of all these severe conditions, failures will happen in field. It is in this context that managers must make maintenance decisions such as: lifetime of potential transformers of different manufactures; probability of failure for a mission equal to concession time; reliability for a mission equal to concession without replacement; reliability for a mission equal to concession with replacement; optimal time of replacement to minimize cost compared to concession time, etc. These problems are addressed in this paper.

The methodology consists in data collection, data munging, exploratory data analysis, variability modelling using probability distributions, forecasting model and economic optimization model. Time-to-failure data (complete, right censured and interval) were registered in an unstructured way over the last 40 years and data extraction involved the use of some specific python codes. The variability of lifetime data is modelled using probability distributions such as normal, Weibull and q-Weibull in some special cases. The forecasting mathematical model is constructed considering the real situation of the power transmission company. The economic model has focus on the optimal replacement model to minimize total cost (maintenance, parts, consequence of failures etc.)

The proposed model is applied to real data and parameters of mathematical models are calibrated appropriately, considering operating time, manufacturers, voltage class, etc. A number of positive contributions have appeared so far: (1) this methodology has allowed maintenance managers to quantify probability of failure for a mission equal to concession time, compare performance of different voltage class; (2) it is important to mention that for some manufacturers the managerial option of replacement will have no effect on reliability whereas for others it will have improvements; (3) for those manufacturers or family of transformers whose failure rate is increasing, if the cost of corrective maintenance is at least 3.5 times the cost of preventive maintenance, the optimal replacement time is after concession time of 30 years; (4) in addition, result supports the replacement decision policy (upon or before failure) and budget planning for the future of the concessionary company.

This article presents part of the results achieved in project PD-06491-0391/2015 (Non-Conventional Mathematical and Computational Models of Quantitative Reliability Focused in Maintenance Problems of Assets in Power Generation and Transmission) which is carried out by FUNDEB and financed by COPEL Power Generation & Transmission under obligation from ANEEL (National Power Energy Agency) Research and Development Program.