應用蒙地卡羅法提升離岸風力發電支撐結構長時域疲勞分析之計算效率

Abstract

本研究以台灣海峽福海風場的離岸風機作為設計目標，建立3.6MW的K型桁架式支撐結構有限元素模型，並且考量離岸風機系統的高度非線性，而採用時間域分析取代頻率域分析。疲勞分析流程參考DNV規範採用熱點應力法，包含銲接效應也納入考量，輔以雨流計數法以及Palmgren-Miner的線性破壞率理論評估疲勞壽命。環境參數根據當地長期氣象統計，建立各參數的聯合機率表。在本文中以三種負載狀況:只施加風負載、只施加波浪負載與兩者都施加，探討疲勞破壞率的統計不確定性。會造成此不確定的因素為時域性負載計算採用隨機過程，這對桁架式支撐結構接點的熱點應力會產生相當大的影響，也意味著需要更大量的模擬數量，已取得能代表該氣象狀況的特徵破壞率。 在現有的設計規範中使用方格法去計算結構的長期破壞率，也就是模擬所有氣象參數所構成的組合，但這也使得模擬數量會隨著考量參數呈現指數成長，若要將上述統計不確定性納入討論，更將造成模擬數量的倍增。為了解決計算效率的問題，使用收斂速度與參數數量無關的蒙地卡羅法，提供了在離岸風機疲勞分析中相當具有潛力的另一個選擇。本研究中藉由10,000個風場、流場與結構疲勞模擬，探討蒙地卡羅法與方格法的計算效率。

The K-type jacket substructure of a 3.6 MW offshore wind turbine, designed for the Fuhai Offshore Wind Farm in Taiwan Strait, modeled in the FEA software. Time-domain analyses were conducted rather than frequency-domain because of the strong nonlinearity of the OWT system. The fatigue assessment procedure in this paper based on the Hot-Spot Stress Approach in accordance with DNV standards, with welding effects considered. Rainflow Counting Algorithm and Palmgren-Miner's linear damage theory were used to predict the fatigue life. The Joint Probability table of long-term statistical environment condition was based on a preliminary site survey. The statistical uncertainty of the distribution of fatigue damage has been assessed by considering 3 different load cases: wind loads only, wave loads only and combination of wind and wave loads. The uncertainty caused by the stochastic processes of wave and wind loads have a great influence on hot-spot stress of welded tubular joints for jacket-type substructure, which means more simulations for each representative short-term environment condition needed to derived the characteristic fatigue damage. Current design standards use Grid-Based method to derive the long-term fatigue damage: scanning over a rectangular grid of metocean inputs, the number of required simulations grows exponentially with dimension, and more simulations needed if we include statistical uncertainty as well. To reduce the number of simulations required, Monte Carlo method used in this paper, sampling simulation condition from the discrete probability distribution in the joint probability table. The theoretical convergence rate of Monte Carlo Method, which is independent of dimension, may offer a potential way to solve the high dimensionality in Offshore Wind Turbine. About 10,000 simulations over the whole joint probability table were discussed to compare the computational efficiency of two sampling method.