Abstract
Wooden utility poles are used in power grids worldwide. Due to the large number of poles, models that can improve the inspection planning and renewal decisions for these components have the potential to yield considerable cost savings. In this paper, a widely used model for the degradation of wood in ground contact, the Timberlife model, is compared with inspection data from about 14,000 Scotch pine utility poles in the Norwegian power grid. This comparison shows that the original Timberlife model substantially underestimates the number of degraded creosote-treated poles relative to the inspection dataset. To address this, we propose an updated degradation model based on the existing Timberlife model and collected inspection data, using a Bayesian workflow. In the updated model, climate (i.e., precipitation and temperature), land type (i.e., forest, agricultural, or built-up area), and preservation treatment (i.e., creosote or CCA) are considered when predicting the degradation rate. Compared with the original Timberlife model, the updated models yield more accurate predictions of the number of degraded poles on a test set of collected inspection data.