29 / 2022-03-02 10:39:13

Enhanced Cable Insulation Resistance Guidelines Based on Weibull Analysis and Diminishing Method

Power cable,insulation resistance,Weibull analysis

Purpose/Aim

The electricity grid company of Singapore uses cable insulation resistance (IR) criteria to assess insulation condition prior to energization. Additional Oscillating Wave Test (OWT) will be conducted to assess state of health of suspected unhealthy medium voltage (MV) power cable circuits. However, there are still cases of MV power cables having passed IR test but failed in service subsequently. Hence, it is essential to better diagnose cable insulation condition with improved IR guidelines to enable timely mitigation measures to be performed.

Experimental/Modeling methods

In this paper, time-to-failure events observed by the company have been firstly analyzed by using a Weibull model. Next, Diminishing Method (DM) has been utilized to compute critical IR values for the guideline. The DM aims to maximize the proportion of failed assets that can be correctly detected and classified (True Positive, TP) while minimizing the number of healthy circuits which are wrongly diagnosed as unhealthy (False Positive, FP) to minimize workloads requiring OWT and follow up. Finally, this IR guidelines for the distribution MV cable networks have been determined by a combination of Weibull and DM.

Results/discussion

By using 2-parameter bended curve of Weibull distribution, three age groups of 22kV and two types (copper and aluminum cores) of 6.6kV power cables have been identified respectively, based on various ageing rates. Critical IR values have been generated for each category by the application of DM then. With the greater value of critical IR, both TP and FP increase. Compromission between TP and FP should be made to obtain the optimal results.

Conclusions

Compared to the currently used IR guidelines, the criteria could potentially recognize a higher number of failed cable circuits albeit at greater workloads. In the future, classifier algorithms based on machine learning will be investigated to decrease the workload further. Details will be provided in the full paper.