| 摘要: | 人類文明發展在帶來進步的同時，亦由於我們對化石能源的過度依賴而產生的大量溫室氣體，引發了全球暖化效應。這種人為氣候變遷不但會助長極端天氣的發生，更可能衝擊生態造成物種滅絕或糧食危機等嚴峻問題。因此，減少溫室氣體排放成為控制全球升溫的一項重要策略。據統計，電力與熱力生產是產生最高碳排放的經濟活動。基於現今社會對能源的需求只將有增無減，利用可再生能源取代傳統發電的做法逐漸被重視，而當中，尤以太陽能發電極速增長成為近年的發展趨勢。當今太陽能市場中最主流的光伏發電方式之核心設備便是我們常聽聞的太陽能板。一個太陽能發電站由大面積的太陽能板組成，並藉此將陽光轉換為可用電力。而太陽能板的裂痕、表面玻璃破碎、熱斑、異物遮擋或污漬覆蓋等都可能影響面板的發電狀況及使用壽命，因此定期檢測是維持其最佳狀態的必要手段。然而，隨著不同形式的太陽能板陸續問世，倚靠人力的傳統維護方式不僅費時、危險且難以實現，如臺灣因建築模式與有限土地等因素，近年以水域型太陽能發電站為發展方向，此便為一不便於以人工巡檢的發電廠種類。因此，近年逐漸成熟的無人機技術成為了巡檢太陽能板的新出路。
綜上所述，本研究針對無人機巡檢水域型太陽能發電站之路徑規劃進行相關研究，自面對現實情境應如何進行問題轉換，至獲得問題網絡圖後應考慮哪些限制建構演算法的流程與方法之探討。文中鎖定巡檢任務中之必行經定位點、充電站位置以及巡檢最佳路徑求解三個議題，建立了一個針對現實情境轉換問題網絡的標準流程，使議題一得以獲得結果，並基於基因演算法提出一調整算法，使其符合本問題情境，可對此無人機路徑問題進行求解，並於結果中得到一組推薦的巡檢任務飛行路徑與推薦之最佳充電站設置位置，完成對議題二、三的討論。最後，本研究在考慮發電廠場域涵蓋面積與分佈狀況、後續影像分析的需求，以及無人機的使用型號三種資訊的結合，並將重疊率視為問題轉換關鍵之下，成功由彰濱工業區之彰濱崙尾東一暨二號電廠轉換出一個具4,798個節點之網絡圖。且於將基因演算法迭代最低限制設為300下，以表現良好的初始解出發，在確保過程中所得解之可行性下，於有限迭代內收斂得到一耗時1,605分鐘之可接受路徑解，以及一個相對穩定的充電站建議設置位置（987.76, 438.62）。
關鍵字：水域型太陽能發電站、光伏發電、光伏檢測、無人航空載具、無人機、無人機路徑規劃。;While the development of human civilization has brought progress, it’s also caused global warming due to the large emissions of GHGs generated by our over-reliance on fossil fuels. This human-induced climate change not only will escalate the occurrence of extreme weather, but also may impact the ecology, causing serious problems such as species extinction or food crisis. Therefore, reducing GHG emissions has become an important strategy to control global warming. According to statistics data, electricity and heat production is the economic sector that produces the highest carbon emissions. As the demand for energy in today′s society will only increase, the methods of using renewable energy to replace traditional power generation have gradually been valued. Among these means, the rapidly growing solar power generation has become a trend in recent years. The solar panel that we often hear is the core equipment of photovoltaic power generation which is the most mainstream generation method in the solar energy market nowadays. A solar power plant consists of large areas of solar panels that convert sunlight into electricity. However, cracks, broken surface glass, hot spots, foreign objects, or stains on the solar panel may affect its generation and life, hence regular inspection is necessary to keep it in its best condition. Nevertheless, with the different types of solar panels coming out, the traditional maintenance method that relies on manual labor is time-consuming, dangerous, and difficult. Thus, drone technology, which has recently matured, has become a new way to inspect solar panels.
To sum up, this research focuses on the path planning for floating solar plant inspection by using drone. Discuss the processes and methods from how to convert problem in realistic situation, to what constraints should be considered to build the algorithm after obtaining the network. The must-be-passed points in the inspection task, the location of the recharging station, and the optimal path for the drone are the three main issues we aim at. For that, we established a standard process for the network conversion of the realistic situation to answer the first issue, and proposed an adjusted algorithm based on Genetic Algorithm which can fit the condition to solve this drone routing problem. After the algorithm, there′s a set of recommended flight paths for inspection task and the recommended optimal recharging station setting positions can be obtained, thus the second and third issues can be completed. Finally, this study considers the combination of three types of information, including the coverage area and distribution of the power plant site, the need for subsequent image analysis, and the use model of drone, and considers the overlap rate as the key to problem conversion. The Zhang Bin solar panel plant has successfully been converted into a network with 4,798 nodes. And under the condition that the minimum limit of Genetic Algorithm iterations set to 300, the algorithm starts from a good initial solution while ensuring the feasibility of the solution obtained in the process, converges into an acceptable path solution that takes 1,605 minutes within a limited iteration, and recommends a relatively stable recharging station position (987.76, 438.62).
Keywords: floating solar plant; solar photovoltaics; photovoltaic inspection; unmanned aerial vehicles (UAV); drone; drone routing problem. |
