| 摘要: | 近年來全球對於發展和應用再生能源越來越關注，尤其是太陽能(Photovoltaics, PV)技術。然而太陽能發電具有高度的間歇性，可能會影響電力系統的電力品質，因此結合電池儲能系統(Battery Energy Storage System, BESS）來提高電網的穩定性成為解決方案之一。但受限於併網法規和儲能系統成本高昂，使用平滑濾波器與BESS相結合來降低成本和穩定輸出功率已成為一種受到廣泛關注的方法。
本文提出的變高斯濾波算法中除了能有效減少時間延遲現象，貼合原功率輸出波型，也擁有較好的平滑程度。此外本文提出一種模糊類神經網路(Fuzzy Neural Network)，透過輸入電池電量、負載使用量以及太陽能功率並模糊化後，獲得各電網的分配比例。本文場域利用儲能系統、太陽能系統與負載整合成三區多微電網系統，結合太陽能平滑並控制統合各區輸出功率，使得系統總體輸出回市電端能有效平滑且符合台灣電力公司規定，透過獲得的分配比例能有效平衡三區微電網的電池電量並使其保持一致。此控制策略除了能有效提高市電端的電力品質，且能維持電量的平衡讓整體電網的使用時間提升。
本文模擬採用MATLAB/Simulink模擬軟體來模擬與比較本文平滑化控制策略的可行性，並使用C語言撰寫相關類神經演算法與平滑算法，最終在實驗場域驗證其控制策略之結果。
;In recent years, there has been increasing global interest in the development and application of renewable energy, particularly Photovoltaics (PV) technology. However, due to the intermittent nature of solar power generation, it may cause problems such as power system faults or blackouts. Combining Battery Energy Storage System (BESS) to improve the stability of the grid has become one of the solutions. However, due to regulatory constraints and the high cost of energy storage systems, combining smoothing filters with BESS to reduce costs and stabilize output power has become a widely recognized method.
This article proposes a variable Gaussian filter algorithm that not only effectively reduces time delays and fits the original power output waveform but also has good smoothness. In addition, a Fuzzy Neural Network is proposed in this article, which inputs the battery power, load usage, and solar power and obtains the allocation ratio of each power grid after fuzzification. The experimental field of this article uses the storage system, PV system, and load integration into three micro-grid systems, combined with solar smoothing and control, to integrate the output power of the three micro-grid systems, enabling effective smoothing and meeting the ramp rate set by Taiwan Power Company. Through the obtained allocation ratio, the battery power of the three micro-grids can be effectively balanced and kept consistent. This control strategy not only effectively improves the power quality at the grid end but also maintains the balance of power, improving the overall use time of the power grid.
This article uses Matlab/Simulink simulation software to simulate and compare the feasibility of the smoothing control strategy proposed in this article. The relevant neural network algorithms and smoothing algorithms are written in C language, and the results of the control strategy are verified in the experimental field. |
