結合類神經網路及灰關聯分析法估測鉛酸蓄電池放電時間

Estimation of the Discharge Time of Lead-Acid Storage Batteries Using Artificial Neural Networks and Grey Correlation Analysis

柯致平(Chih-Ping Ke)

指導教授 ： 龍仁光

國立臺北科技大學/電資學院/電機工程系所/碩士(2010年)

https://doi.org/10.6841/NTUT.2010.00108

摘要

通信用電力是使用交換式直流供電設備（Switching Mode Rectifier Power Supply：簡稱SMRPS）將交流電源轉換成直流電源供給通信設備使用，每套SMRPS依容量都配有足夠之閥調式鉛酸蓄電池（Valve Requlated Lead-Acid Storage Battery：簡稱VRLA蓄電池），其用途為交流電源中斷時供給設備所需之直流電源，如蓄電池無法供應電力時，則會造成通信中斷。目前對蓄電池容量測試方法為放電容量測試法(統稱安培小時法)，可以準確判斷蓄電池之容量，但耗能及浪費人力，且作業時間長，亦會影響通信系統之穩定性。對於蓄電池放電時間估測的方法，本論文是以各種不同放電電流對蓄電池進行放電容量測試，將所測得之端電壓及電流數據，輸入至倒傳遞類神經網路，經過訓練學習後，用以估測蓄電池之放電時間，日後用監控系統隨時監測蓄電池相關資料，利用定期發電機測試時取得相關數據加以估測蓄電池放電時間，並使用灰關聯分析法判斷測試結果，本方法具有縮短作業時間（放電及充電）約60~1601倍、減少人力資源、節省能源約60~1360倍之優點，由實例證明，應用倒傳遞類神經網路與灰關聯分析法，訓練時含放電電流可將均方誤差收斂至0.0405%以內，因此，上述方式應用在蓄電池放電時間估測方面是有不錯的成果。

關鍵字

倒傳遞類神經網路 ； 灰關聯分析法 ； 鉛酸蓄電池 ； 放電時間

並列摘要

The power that we use in communication is “Switching Mode Rectifier Power Supply (Abbreviation: SMRPS)”, it transfers AC power to DC power then supplies for communication equipment, depends on the capacity, each SMRPS can have enough Valve Regulated Lead-Acid Storage Battery (Abbreviation: VRLA Battery), the function is when the raw power cuts, it can supply the equipment, if the battery cannot supply the power, communication will fail. So far, the method of testing storage battery is discharge capacity testing. By this way, we could determine the capacity exactly. But it would waste resource, manpower, working time and affect the stability of communication system. The way of value discharge time is releasing different powers to the storage battery. And then put the numbers of voltage and current to Back-propagation Artificial Neural Network. After learning and training, it could be used to test the discharge time of variety power. Then, detecting the related data of discharge battery by remote monitor system. Getting the number by using routine generator to estimate discharge time. This method shortens the working hours approximately 60~1601 times, reduces manpower and saves energies approximately 61~1360 times determined by Grey Correlation Analysis. In this case, application of Back-propagation Artificial Neural Network and Grey Correlation Analysis, when training contains the discharge current to be possible mean error restraining to 0.0405%, the above way application in the battery discharge time estimated that the aspect has the good achievement.

並列關鍵字

Back-propagation Artificial Neural Network ； Grey Corrlation Analusis ； Lead-Acid Storage Battery ； Discharge Time

參考文獻

[36]羅強華，類神經網路-MATLAB的應用，清蔚科技股份有限公司，2001。

[5]Grewal S.,Grant D.A., 〝A Novel Technique For Modeling The State of Charge of Lithium Ion Batteries Using Artificial Neural Network,〞IEE Conference Publication, No.484. , 2001, pp.174-179。

[16]EXIDE, GNB INDUSTARIAL POWER SECTION 62.61, 2008。

[17]J.H. Aylor, A. Thieme and B.W. Johnso, 〝A battery state-of-charge indicator for electric wheelchairs,〞 IEEE Trans. on Industrial Electronics, Vol. 39, 1992, pp.398-409.

[18]Sul Seung-Ki and Lee Sang-Joon, 〝An integral battery charger for four-wheel drive electric vehicle,〞 IEEE Trans. on Industry Applications, Vol. 31, 1995, pp.1096-1099.