應用免疫-田口混合演算法於電力系統低頻卸載最佳規劃

Optimal Planning of Power System Underfrequency Load Shedding Using Hybrid Taguchi-Immune Algorithm

劉正善(Jeng-Shang Liu)

指導教授 ： 陳昭榮

國立臺北科技大學/電資學院/電機工程系所/碩士(2011年)

摘要

台電為獨立的電力系統，不若北美區域互聯系統穩定，頻率容易受到干擾，情況屬輕微，尚可由自動發電控制或緊急發電系統回復至穩定。惟如當發生重大事故時，例如電廠發電機組跳脫或電力系統解聯，負載供需嚴重失衡會使頻率驟降，不僅會損害電力相關設備，亦會影響用戶供電品質。發電機在低頻下運轉除了會影響葉片使用壽命，若此時不能執行適當的卸載，恐將引起發電機組連鎖跳脫，造成全停電。
本論文研究的目的主要是鑑於台電系統使用固定頻率式低頻電驛來設定卸載，惟目前新型電驛均具有偵測頻率變化率功能，因此，本論文使用固定頻率及具有變動頻率變化率功能低頻保護電驛作為系統頻率偵測及動作元件，以此控制斷路器跳脫負載。另在模擬方法的選擇上，傳統是利用單機無限匯流排系統，本論文採用美國API公司所研發PSS/E套裝軟體，改變以往作法，大幅提高實驗結果的精確度。
本論文提出免疫-田口混合演算法，以台電2010年5月尖峰及2014年冬季離峰(預估核四1部機商轉)電力系統為例，假設台電系統可能發生多個不同嚴重程度的重大事故，造成系統頻率驟降，此時採取卸載策略使系統回復穩定，保護系統於安全狀態兼顧停電量最小化目標。此進化之演算法特點，是在免疫演算法交配與突變間，插入田口式直交表實驗產生可能後代解，在實驗過程中如有比菁英更佳解並加以取代，求解低頻電驛設定卸載頻率、卸載量及頻率變化率等參數之全域最佳解。實驗結果免疫-田口混合演算法與免疫演算法或台電規劃設定相較，證明能以較短的時間更接近最佳解，此方法確實可以得到較低卸載量仍能使系統恢復正常運轉。本論文提出的方法，可提供任何電力公司作為訂定每年之卸載策略。

關鍵字

免疫演算法 ； 免疫-田口混合演算法 ； 發電機組跳脫 ； 低頻卸載

並列摘要

Taipower system, unlike the interconnected system in North America region being quite stable, is independent thus its frequency can easily be interfered. When the incident is minor, the power system can be stabilized with automatic power control or emergency power system. However, when there is major incident such as power plant generator unit tripping or power system tie line tripping, serious imbalance between loads of supply-and-demand will result in rapid drop of frequency. This would not only damage related power devices, but also affect the quality of power supplied to users. When generator is operated in low frequency, not only service life of the blades may be affected, when no appropriate load shedding is implemented in time, but it could also result in sequence of generator unit tripping and cause total blackout.
Taipower system uses fixed-frequency-type low frequency relay for load shedding. On the other hand, all current new type reply has the function to detect frequency change rate. Therefore, for the purpose of controlling load shedding of circuit breaker tripping, the thesis uses fixed-frequency-type low frequency protection relay and low frequency protection relay with detective function of frequency change as system frequency detecting and acting elements. For selection of simulation method, the thesis adopts PSS/E software package by American API rather than traditional OMIB (One-Machine-Infinite-Bus) system to greatly improve precision of experimental result.
Thesis uses immune-Taguchi hybrid algorithm and taking power system of Taipower system in high peak May 2010 and off-peak winter 2014 (estimated operation of Lungmen Machine #1) as model. It is assumed Taipower system may experience multiple incidents with different level of severity which result in rapid drop of system frequency. The goal is to adopt load shedding strategy to stabilize the system and protect system in a safe state with minimal blackout. The characteristic of this evolutionary algorithm is to insert Taguchi orthogonal array between crossover and mutation to produce possible descendant. During the experiment, if a better solution exists, such solution will replace former solution to find global extreme value for parameters such as load shedding frequency of low frequency relay, volume of load shedding, and frequency change rate. Comparing Hybrid Immune-Taguchi Algorithm with Immune Algorithm or Taipower system setting, it is proved that Hybrid Immune-Taguchi Algorithm finds the best solution in shortest time; and the solution definitely uses lower load shedding volume to resume the system to normal. The method proposed by this thesis may provide any power companies to set up annual load shedding strategy.

並列關鍵字

Immune Algorithm ； Hybrid Immune- Taguchi Algorithm ； Generator unit trip ； Underfrequency load Shedding

參考文獻

[38] 夏倬彥，應用免疫演算法於輸電系統最佳化無效功率調度，碩士論文，國立台北科技大學電機工程研究所，2008年。

[3] S. Lindahl, G. Runvik, G. Stranne, “Operational experience of load shedding and new requirements on frequency relays,” Sixth International Conference on Developments in Power System Protection, pp. 262-265, March 1997.

[4] J. R. Jones, W. D. Kirkland, “Computer algorithm for selection of frequency relays for load shedding,” IEEE Computer Applications in Power Systems, pp. 21–25, Jan. 1988.

[5] C. Concordia, L. H. Fink, G. Poullikkas, “Load shedding on an isolated system,” IEEE Transactions on Power Systems, pp. 1467–1472, Oct. 1995.

[7] J. G. Thompson, B. Fox, “Adaptive load shedding for isolated power systems,” IEEE Proceedings Generation Transmission and Distribution, pp. 491-496, Sep. 1994.