基於WRF–SWAN耦合與JONSWAP波譜之離岸風電場極端風浪特性分析

Abstract

臺灣海峽具有良好風能資源，但每年夏秋兩季受颱風頻繁侵襲，極端風浪事件對離岸風機結構安全性與營運維護造成重大挑戰。本研究以2001年至2023年間36起侵臺颱風作為研究案例，利用天氣研究和預報WRF (Weather Research and Forecasting)模式與第三代風浪模擬模型SWAN (Simulating WAves Nearshore)耦合模擬，並透過浮標資料驗證。模擬結果顯示，WRF能有效重現颱風風場之變化趨勢，但當颱風結構受地形干擾時，模擬結果會產生誤差。第3類路徑及第7類路徑之風速模擬誤差較大，第3類路徑風速均方根誤差平均達6.21 m/s，第7類路徑相關係數平均僅0.24，原因主要來自颱風中心接近陸地與地形遮蔽效應所致。波場方面，SWAN能有效模擬示性波高、平均週期與波向之整體變化趨勢。然而，波高模擬之準確性高度依賴WRF提供之風場，因此WRF模擬誤差會進一步放大SWAN波高模擬誤差。第2類路徑波高均方根誤差平均為1.04公尺，第7類路徑波高相關係數平均僅0.16，兩類路徑誤差較大。第6類路徑模擬波高均方根誤差平均僅0.35公尺，相關係數平均為0.83，模擬誤差小。本研究亦將新竹浮標處之一維波譜以JONSWAP波譜擬合，整體波譜平均飛利浦常數與尖峯集中因子分別為0.0081與1.11。分析另外五座離岸風場極端波況下之波譜差異。結果顯示，近岸之允能風場波浪能量分散，尖峯集中因子較低，平均為1.09。彰芳西島風場則具有最高之尖峯集中因子平均值1.29，波能較高且集中，極端波況下波浪對風力發電機影響最大。整體而言，WRF耦合SWAN模擬能重現颱風侵襲時臺灣鄰近海域之風浪，並進一步提供不同風場面臨極端波況之JONSWAP波譜參數，作為離岸風機結構設計與運維之參考依據。

The Taiwan Strait possesses abundant wind energy resources. However, frequent typhoon occurrences in summer and autumn pose significant challenges to the structural safety and operational maintenance of offshore wind turbines. This study analyzed 36 typhoon cases between 2001 and 2023, utilizing a coupled numerical model integrating the Weather Research and Forecasting (WRF) model with the third-generation wave simulation model, Simulating WAves Nearshore (SWAN), and validated the simulation results against observational data. Simulation results demonstrated that WRF could capture the trends of wind fields of typhoon. Notable discrepancies arose when typhoons were disrupted by complex terrain interactions. The largest simulation errors occurred in the typhoon paths classified as category 3 and category 7. The average root mean square error (RMSE) for wind speed simulations under category 3 reached 6.21 m/s, while the average correlation coefficient for category 7 was only 0.24. These errors primarily resulted from proximity of typhoon centers to land and terrain-induced shielding effects. Wave simulations indicated that SWAN effectively reproduced the trends of wave fields. However, the accuracy of simulated wave heights heavily depended on the accuracy of wind fields generated by WRF. The average RMSE for simulated wave heights in category 2 was 1.04 m, with an average correlation coefficient of merely 0.16 for category 7. Conversely, category 6 exhibited smaller errors, with an average RMSE of 0.35 m and an average correlation coefficient of 0.83. Additionally, the research fitted the one-dimensional wave spectrum with the JONSWAP spectrum, determining an overall average Phillips constant of 0.0081 and a peak enhancement factor of 1.11 at Hsinchu buoy. The study also analyzed the wave spectra under extreme conditions at five offshore wind farm sites around Taiwan. Results indicated that nearshore wind farms such as the Yunlin site exhibited lower peak enhancement factors averaging 1.09. Situated in Taiwan’s central offshore region, the Changfang and Xidao wind farms exhibited the highest mean peak-enhancement factor of 1.29, signifying more intense and concentrated wave energy. Accordingly, extreme sea states are expected to impose the greatest wave loading on turbines at these sites. In conclusion, the coupled WRF-SWAN simulation approach employed in this study reliably models wind and wave conditions in Taiwan during typhoon events. Furthermore, it provides detailed JONSWAP spectrum parameters for extreme wave conditions at various offshore wind farms, serving as critical references for offshore wind turbine design and operational planning.