TRANSMISSION LINE ICE ACCRETION PROBABILITY PREDICTION AND DE-ICING THRESHOLD OPTIMIZATION BASED ON BAYESIAN EXTREME VALUE THEORY AND MONTE CARLO SIMULATION

Abstract

Transmission line icing is a rare and severe hazard; therefore, a fixed de-icing threshold cannot account for forecast uncertainty, accumulated long-term risk, and the economic trade-off between de-icing costs and tower collapse losses. Based on 30 years of annual maximum ice-thickness data, real-time weather forecasts and operating cost parameters, this paper builds a probabilistic prediction and de-icing decision model. A generalised extreme value distribution is used to obtain the prior via maximum likelihood estimation, Bayesian updating and Markov chain Monte Carlo sampling generate the posterior distribution, a homogeneous Poisson process determines the long-term event frequency, and finally, Bayesian expected loss yields a dynamic de-icing threshold. The fitted GEV parameters are μ=11.95, σ=3.08, ξ=-0.19, and the 50-year and 100-year return levels are 21.5 and 23.9 mm, respectively. Under the given severe-weather condition, the posterior mean is 15.6 mm, and the probability of exceeding 15 mm is 0.84. The probabilities of at least one event within 10, 20, 30 and 50 years are 0.9030, 0.9906, 0.9991 and 0.9999, respectively. The optimal de-icing threshold ranges from 14.2 to 18.6 mm, reducing the expected total cost by about 18.7% compared with a fixed-threshold policy. This report provides verifiable statistical decision support for the construction of a high-probability warning system and proactive de-icing measures for transmission lines.