臺灣博碩士論文加值系統

本文利用量子類神經網路學習方式進行短期負載預測。利用量子類神經網路特性應用於電力負載訊號上，可藉著自我學習推衍負載量與影響因素之間的映射關係，進而建構模型預測非穩定電力負載訊號。較大的學習速率 具有較大的網路連接權重修正值，可以加速網路學習，但也容易造成過度修正，而使得網路震盪無法收斂，較小的學習速率可避免震盪，卻可能減緩網路的收斂速度。同時，收斂於局部最佳解的問題亦常存在於現有之類神經網路訓練方法。為了加速網路的收斂速度及避免震盪現象，通常在誤差回傳時加入一慣動量α。
本研究針對非穩定電力負載預測，提出修正型量子類神經網路，也就是針對類神經網路學習陷入局部最小值的問題，利用模糊推論的方式找出學習速率η與慣動量α的最佳組合，針對修正型量子類神經網路效能的參數加以改善。實例顯示，使用修正型量子類神經網路應用於台電公司負載預測時，較不易陷入局部最小值，預測平均百分比誤差值也明顯的比現有倒傳遞類神經網路為佳。

This thesis presents a fuzzy based quantum neural networks method for short-term load forecasting. By using the nature of the quantum neural network. the relationship between the load and the influencing factors can be derived through self-learning to establish a forecasting model for non-stationary power load signals. With larger the learning rate , we have larger the correction values of the interconnecting weights to speed up the learning. However, it may also cause excessive correction and lead to oscillation and divergence. With a smaller learning rate, the problem of oscillation can be avoided, but the speed of convergence would be slowed. Also, the problem of trapping at the local optimum exist in training the neural networks. In order to speed up the convergence of the network and avoid the oscillation problem, a momentum coefficient during return of error signal is added.
The research is aimed at forecasting the non-stationary power load through modified quantum neural networks. The optimal combination of learning rate and momentum coefficient is obtained by using fuzzy inference to overcome the local optimum problem. The impact of training parameters on performance of the neural networks is investigated. Numerical results obtained from practical Taipower load forecasting reveal that not only can the local optimum problem be avoided, but the average forecasting accuracy of the proposed method is shown to be superior to the existing back-propagation neural networks.

目錄
中文摘要…………………………………………………………………………I
英文摘要……………………………………………………………………II
致謝………………………………………………………………………...III
目錄………………………………………………………………………...IV
圖目錄…………………………………………………………………….VII
表目錄……………………………………………………………………...IX

第一章 緒論……………………………………………………………….1
1.1 前言…………………………………………………………………………1
1.2 文獻回顧…………………………………………………………………...2
1.3 研究目的…………………………………………………………………...5
1.4 論文架構…………………………………………………………………...7

第二章 類神經網路與學習理論………………………………………….9
2.1 簡介…………………………………………………………………………9
2.2 類神經網路的特色與優點…………………………………………...11
2.3 類神經網路的學習方式………………………………………..12
2.4 類神經網路架構………………………………………………..15
2.4.1 前向式類神經網路………………………………………15
2.4.2 回饋式人工神經網路……………………………………16
2.5網路參數的選擇………………………………………………………....17
2.6網路學習法則……………………………………………………………19
2.7過度學習效應…………………………………………….……..21

第三章 量子類神經網路……………………………………………………….. 22
3.1量子類神經網路之架構………………………………………...22
3.1.1 量子區間與跳躍點……………………………………...25
3.2 量子類神經網路之學習法則……………………..………….......27
3.2.1 類神經網路之權值修正………………………………………27
3.2.2 量子區間更新…………………………………………………..29
3.2.3 量子演算法……………………………………………………...30
3.3 模糊推論(Fuzzy Inference) …………………………………………..35
3.3.1 模糊條件語句…………………………………………………..35
3.4 改良型學習演算法……………………………………………………..36
3.4.1 倒傳遞學習演算法結合模糊推論………………………….36

第四章 負載預測數值結果…………………………………………………….41
4.1 簡介…………………………………………………...…………………..41
4.2 輸入變量與負載之關係…………………………………...………….42
4.3 訓練及測試集合的選擇…………………………….………………...42
4.4 每日尖峰負載預測…………………………………..…..……….44
4.5 未來一小時的負載預測………………………………..………57
4.6 台電負載預測結果分析..............................................................62

第五章 結論與未來研究方向..............................................................................65
5.1 結論…………………………………………………………………….....65
5.2 未來研究方向...........................................................................................65
附錄A ……………………………………………………………………………….67參考文獻……………………………………………………………………………….70

圖目錄
圖 1-1 類神經網路學習誤差與權值關係……………………………….7
圖 2-1 神 經 元 模 型…………………………………………………10
圖 2-2 Sigmoid 轉移曲線………………………………………………..11
圖 2-3 前向式類神經網路架……………………………………………15
圖 2-4 回饋式神經網路架構………………………….………………...16
圖 3-1 量子類神經網路架構…………………………………………....23
圖 3-2 的位置對量子區間的影響………….......................................25
圖 3-3 重疊對量子區間的影響………...............................................26
圖 3-4 調整s型作用函數的斜率之影響比較………………………….26
圖 3-5 誤差百分比( )歸屬函數…………..........................................38
圖 3-6 誤差變化量百分比(Δ )歸屬函數…………..........................38
圖 3-7 慣動量( )歸屬函數………….....................................................39
圖 3-8 學習速率(η)歸屬函數……………............................................39
圖 4-1 每日尖峰負載預測神經網路架構………….…………………...46
圖 4-2 台電民國87年春季尖峰負載曲線………………………………49
圖 4-3 台電民國87年夏季尖峰負載曲線………………………………50
圖 4-4 台電民國87年秋季尖峰負載曲線………………………………51
圖 4-5 台電民國87年冬季尖峰負載曲線………………………………52
圖 4-6 台電民國88年春季尖峰負載曲線………………………………53
圖 4-7 台電民國88年夏季尖峰負載曲線………………………………54
圖 4-8 台電民國88年秋季尖峰負載曲線………………………………55
圖 4-9 台電民國88年冬季尖峰負載曲線………………………………56
圖 4-10 未來ㄧ小時的負載預測神經網路架構…………………………58

表目錄
表 3-1 描述慣動量 模糊邏輯規則表......................................................40
表 3-2 描述學習速率η模糊邏輯規則表………………………..…...….40
表 4-1 每日尖峰負載預測使用之網路參數………………...…………...47
表 4-2未來ㄧ小時的負載預測之網路參數………………. …………….59
表 4-3 台電民國87年未來一小時負載資料模擬預測表………………..60
表 4-4 台電民國88年未來一小時負載資料模擬預測表………………..61

參考文獻
[1] 張夢麟、李勇昇，”利用類神經網路作短期性負載預測”，82年技術研討會論文集，1993年11月.
[2] M.T.Hangn and S.M.Behr, “The time series approach to short-term
load forecasting,” IEEE Trans. on Power Apparatus and Systems,
Vol.2, No.3, pp.785-791, 1987.
[3] G.T. Heineman, D.A. Nordman, and E. C. plant, “The Relationship Between Summer Weather and Summer Loads, A Regression Analysis,”IEEE Trans. Power App. System, Vol. PAS85, No.11,pp.1144-154,1996.
[4] 溫坤禮、黃宜豐、李元秉，灰預測原理與應用，全華科技圖書股份有限公司，1991年.
[5] 葉怡成，應用神經網路，儒林圖書，1997年７月.
[6] R. C. Hwang, H. C. Huang, J. G. Hsieh,“A New Artificial Intelligent Peak Power Load Forecaster Based On Non-Fixed Neural Networks”, International Journal of Electrical Power and Energy Systems, Vol. 24, Issue: 3, pp. 245–250, 2002 .
[7] R. C. Hwang, H. C. Huang, Y. J. Chen, J. G. Hsieh,“Power Load
Forecasting By Neural Network With A New Learning Process For Considering Overtraining Problem”, 1998 International Conference on Energy Management and Power Delivery, Vol. 1, pp. 317–322, 3–5 .1998.
[8] P. K Dash, A. C Liew. and S.Rahman, “Peak load forecasting using
a fuzzy neural network,” Electric Power Systems Research, Vol. 32,
No. 1, 1995, pp. 19-23.
[9] 葉怡成，類神經網路模式應用與實作，儒林圖書，1993年 1 月
[10] 張維哲，人工神經網路，全欣資訊圖書，1992年10月.
[11] 中國科學技術大學生物醫學工程跨系委員會編著，神經網路及其應用，儒林圖書，1993年1月。
[12] H.M.Al-Hamadi, and S.A. Soliman, “Short-term electric load forecasting based on Kalman filtering algorithm with moving window weather and load model,” Electric Power Systems Research., Vol. 68, No. 1, 2004, pp. 47 – 59.
[13] P.F. Pai, and W.C .Hong, “Support vector machines with simulated annealing algorithms in electricity load forecasting” Energy Conversion and Management , Vol. 46, No. 17, 2005, pp. 2669-2688.
[14] 王進德、蕭大全編著，類神經網路與模糊控制理論入門，全華科技圖書股份有限公司，1994年9月.
[15] T.Yalcinoz and U.Eminoglu, “Short term and medium term power distribution load forecasting by neural networks,” Energy Conversion and Management, Vol. 46, No. 9-10, 2005, pp. 1393-1405.
[16] 周政宏編著，神經網路理論與實務，松岡電腦圖書資料股份有限公司
[17] 傅心家編著，神經網路導論，第三波文化事業股份有限公司，1991年9月.
[18] R. C. Hwang, H. C. Huang, J. G. Hsieh,“Short-Term Power Load Forecasting By Neural Network With Stochastic Back-Propagation Learning Algorithm”, PowerEngineering Society Winter Meeting, 2000. IEEE, Vol. 3, pp. 1790–1795, 23–27 . 2000.
[19] S. J.Kiartzis, “Short-term load forecasting using neural networks,” Fuel and Energy Abstracts ,Vol. 36, No. 6, 1995, pp. 434.
[20] N. Matsui, N. Kouda, H. Nishimura,“Neural Network Based On QBP And Its Performance”, Proceedings of the IEEE-INNS-ENNS International Joint Conference on Neural Networks, 2000, Vol. 3 , pp. 247–252, 24–27 .2000.
[21] 王俊波、譚永東編著，神經網路與神經計算機原理與應用，格致圖書公司，1992年10月。
[22] J. Zhou,“Automatic Detection Of Premature Ventricular Contraction Using Quantum Neural Networks”, Third IEEE Symposium on Bioinformatics and Bioengineering, pp. 169–173, 10–12 2003
[23] G. Purushothaman, N. B. Karayiannis,“Quantum Neural Networks (QNNs):Inherently Fuzzy Feedforward Neural Networks”, IEEE Transactions on Neural Networks, Vol. 8, Issue: 3, pp. 679–693, 1997
[24] F. Li, S. Zhao, B. Zheng,“Quantum Neural Network In Speech Recognition”,2002 6th International Conference on Signal Processing, Vol. 2, pp. 1267–1270,26–30. 2002
[25] R. Kretzschmar, R. Bueler, N. B. Karayiannis, F. Eggimann,“Quantum Neural Networks Versus Conventional Feedforward Neural Networks: An Experimental Study”, Proceedings of the 2000 IEEE Signal Processing Society Workshop, Neural Networks for Signal Processing, Vol. 1, pp. 328–337, 11–13. 2000
[26] A.M. Sharaf, and T.L. Tjing, “A novel neuro-fuzzy based self-correcting online electric load forecasting model,” Electric Power Systems Research ,Vol. 34, No. 2, 1995, pp. 121-125
[27] V. Miranda and C. Monteiro, “Fuzzy inference in spatial load forecasting,” Proceedings of 2000 IEEE/PES Winter Meeting, Vol.2, pp.1063-1068, 2000.
[28] R. H. Liang, and C. C. Cheng,“Short-Term Load Forecasting By A Neuro-Fuzzy Based Approach”, International Journal of Electrical Power and Energy Systems, Vol. 24, Issue: 2, pp. 103—111, February, 2002