https://scholars.lib.ntu.edu.tw/handle/123456789/86278
標題: | 風力發電機發電量之推估 Estimation of Energy Output from Wind Energy Conversion Systems |
作者: | 杜逸龍 Tu, Yi-Long |
關鍵字: | 再生能源;風力發電;韋伯分佈;時間序列;標準性能曲線;類神經網路;Renewable energy;Wind power;Weibull distribution;Time series;Nominal performance curve;Artificial Neuron Nnetwork | 公開日期: | 2009 | 摘要: | 自18世紀人類大量使用石化燃料,造成溫室效應的惡化與氣候的變遷,迫使人們開始尋找不具污染性的再生能源來替代石化燃料。在過去10年,風力發電比起其他再生能源的發展來得快速且穩定,而在進行某一地區的風能開發時,若能確實掌握該區域發電量(或負載因子),可以減少風力發電開發廠商對該區域風力發電量之疑慮並減少其投資風險。推估風機發電量的傳統方法是直接利用風機廠商提供的標準性能曲線 (Nominal Performance Curve),再透過其中風速與發電量的關係推算發電量,這種方法可採「韋伯方式」或「時間序列方式」估算其發電量,而因這兩種推估方式的準確性較少被世界各國的文獻所探討,本文自第三章開始即利用最早運轉於台灣地區的兩座風力電廠 (麥寮發電廠與中屯發電廠) 2002年至2006年風機發電的資料,進行這兩種推估方式的準確性之探討與比較。結果發現兩種方法再推估月發電時都有可接受的準確性,其中時間序列的方法較韋伯分佈的方法為佳。者,雖然傳統方法推估發電量有簡單方便而且直接的優點,但是因為風機的標準性能曲線係風機廠商將風機控制在固定條件下 (以Vestas47-660kW的風機為例,其控制條件是在空氣密度1.225 kg/m3、10%紊流以及採旋角控制的扇葉) 的風場所訂定,故一旦風機所處的風場環境與訂定標準性能曲線不同時,極有可能出現推估發電量不準的情形,尤其在台灣,因其受到季風氣候的影響,會有強弱風期之明顯差別,此種推估的誤差會更明顯。因此,為改善此推估發電量的缺點,本文第四章也利用風機運轉的歷史資料,結合類神經網路,探討不同長與不同類型度資料庫推估發電功率的準確性,結果發現在相同資料庫長度的情形下,以各半年的強弱風期所推估的結果最佳,在麥寮約在3.1%至3.8% 之間,在中屯約為1.6%至1.8%。本文第五章則探討以前一季及前一旬的不同輸入值,如風速、發電功率、風機的迎風轉角 (yaw angle) 或扇葉旋角 (pitch angle),對發電功率推估準確性的影響,結果發現以風速搭配前筆發電功率的推估模式最佳。 The development of wind energy has rapidly and steadily progressed then other renewable energy for the last decade, which is driven by global warming and weather change. An accurate estimation of wind energy output (capacity factor) of a wind energy conversion system (WECS) of a site can help producer to reduce the risk of loss from investment of wind energy. Traditional approach is using the manufacturer’s nominal performance curve to estimate energy output. The Weibull method and chronological (time-series) method are usually used to collocate nominal performance curve to estimate wind energy output. Due to the lack of academic work that compares the differences of the measured and calculated capacity factors of WECS, the wind speed and wind power output data from 2002 to 2006 of the wind power stations, located in Mailiao and Jhongtun, Taiwan, are used to further explore the advantages and drawbacks of using these two methods for estimating capacity factors of WECS in chapter 3. It is shown that thecapacity factors calculated from the time-series approach have better agreement with the actual capacity factors than the Weibull approach.he traditional approach is simple and direct for estimating energy output, but it would cause significant estimation errors. Taking the Vestas V47-660kW as example, the performance curve is made under the standard condition of 1.225 kg/m3 air density, 15℃ temperature and 10% turbulence intensity. The estimation of energy output would result in significant errors, if the wind environment is different from the condition of the nominal performance curve. In addition, because of the climatic features of Asia monsoon, there are different prominent wind periods during a year in Taiwan. For improving the drawbacks of the traditional approach, artificial neuron network (ANN) was chosen to estimate wind energy output. The accuracy of estimating wind energy output by different dataase lengths and different types of database for ANN was studied in chapter 4. The results show that the type of strong and weak wind periods database have the best agreement with the actual capacity factors of all the database types with the same dataase lengths. he advantages and flaws of using different input variable for ANN, such as current and previous observations of wind speed, power output, pitch angle or yaw angle for the previous season and a period of ten days were investigated in chapter 5. It is shown that the ANN model with current observation of wind speed and previous observation of power output yields the best performance. |
URI: | http://ntur.lib.ntu.edu.tw//handle/246246/181151 |
顯示於: | 生物環境系統工程學系 |
檔案 | 描述 | 大小 | 格式 | |
---|---|---|---|---|
ntu-98-D93622006-1.pdf | 23.53 kB | Adobe PDF | 檢視/開啟 |
在 IR 系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。