基于LIESN的光伏功率預測研究
孫鵬1,張依強1,焦程煒2
(1 國網山東省電力公司菏澤供電公司,山東 菏澤 274000;2 國網山東省電力公司萊蕪供電公司,山東 萊蕪 271100)
摘 要:為了光伏功率預測結果有更好的準確性與普適性,提出基于泄漏積分型回聲狀態(tài)網絡(LIESN) 的具有在線學習功能的預測方法。在回聲狀態(tài)網絡(ESN) 中引入泄漏積分型神經元,增強儲備池的短期記憶能力;分析了LIESN的參數對其光伏功率預測性能的影響,得到優(yōu)化后的預測模型;利用最小二乘在線學習算法對模型實施訓練,得到最終的在線學習LIESN預測模型。實例證明,該算法可完成復雜的建模且適用于多種天氣情況,預測精度優(yōu)于BP神經網絡、經典ESN及LIESN模型,驗證了方法的有效性。
關鍵詞:回聲狀態(tài)網絡;泄漏積分;神經元;光伏功率預測;在線學習
中圖分類號:TM615 文獻標識碼:A 文章編號:1007-3175(2018)04-0018-06
Online-Learning PV Power Forecasting Based on Leaky-Integrator ESN
SUN Peng1, ZHANG Yi-qiang1, JIAO Cheng-wei2
(1 Heze Power Supply Company, Heze 274000, China; 2 Laiwu Power Supply Company, Laiwu 2711 00, China)
Abstract: In order to enhance computing accuracy and universality of photovoltaic (PV) power forecasting, this paper proposed a online-learning method based on leaky-integrator echo state network(LIESN). Leaky-integrator neurons were introduced to plain ESN and the short-term memory ability was promoted. The impact of parameters of LIESN on PV power forecasting performance was analyzed and an optimized model was obtained. The model was trained by least squares online learning algorithm and final forecasting was obtained. By practical examples, complicated model can be established and applied to various weather conditions. The forecasting accuracy was superior to the BP neural network and plain ESN and the validity of proposed method is testified.
Key words: echo state network; leaky-integarator; neurons; photovoltaic power forecasting; online learning
參考文獻
[1] FERNANDEZ-JIMENEZ L A, MUNOZ-JIMENEZ A, FALCES A, et al. Short-term power forecasting system for photovoltaic plants[J]. Renewable Energy,2012,44(4):311-317.
[2] 龔鶯飛,魯宗相,喬穎,等. 光伏功率預測技術[J]. 電力系統(tǒng)自動化,2016,40(4):140-151.
[3] 陳昌松,段善旭,殷進軍. 基于神經網絡的光伏陣列發(fā)電預測模型的設計[J]. 電工技術學報,2009,24(9):153-158.
[4] 李鵬梅,臧傳治,王侃侃. 基于相似日和神經網絡的光伏發(fā)電預測[J]. 可再生能源,2013,31(10):1-4.
[5] 黃磊,舒杰,姜桂秀,等. 基于多維時間序列局部支持向量回歸的微網光伏發(fā)電預測[J]. 電力系統(tǒng)自動化,2014,38(5):19-24.
[6] SHI J, LEE W J, LIU Y, et al.Forecasting Power Output of Photovoltaic Systems Based on Weather Classification and Support Vector Machines[J].IEEE Transactions on Industry Applications,2012,48(3):1064-1069.
[7] LI Y T, SU Y, SHU L J.An ARMAX model for forecasting the power output of a grid connected photovoltaic system[J].Renewable Energy,2014,66(6):78-89.
[8] JAYAWARDENE I, VENAYAGAMOORTHY G K.Comparisonof echo state network and extreme learning machine for PV power prediction[C]//2014 IEEE Symposium on Computational Intelligence Applications in Smart Grid (CIASG),2014.
[9] JAEGER H. The “echo state” approach to analysing and training recurrent neural networks[J].Uberwachtes Lernen,2001:1-47.
[10] JAEGER H, LUKOSEVICIUS M, DAN P, et al. Optimization and applications of echo state networks with leaky-integrator neurons[J]. Neural Networks,2007,20(3):335-352.
[11] 羅熊,黎江,孫增圻. 回聲狀態(tài)網絡的研究進展[J].北京科技大學學報,2012,34(2):217-222.
[12] 彭宇,王建民,彭喜元. 基于回聲狀態(tài)網絡的時間序列預測方法研究[J]. 電子學報,2010,38(B2):148-154.
[13] 倫淑嫻,林健,姚顯雙. 基于小世界回聲狀態(tài)網的時間序列預測[J]. 自動化學報,2015,41(9):1669-1679.
[14] LUN S X, YAO X S, QI H Y, HU H F. A novel model of leaky integrator echo state network for time-series prediction[J].Neurocomputing,2015,159(1):58-66.
[15] 李軍,岳文琦. 基于泄漏積分型回聲狀態(tài)網絡的軟測量動態(tài)建模方法及應用[J]. 化工學報,2014,65(10):4004-4014.
[16] 張浩然, 汪曉東. 回歸最小二乘支持向量機的增量和在線式學習算法[J]. 計算機學報,2006,29(3):400-406.
[17] 郭振凱,宋召青,毛劍琴. 一種改進的在線最小二乘支持向量機回歸算法[J]. 控制與決策,2009,24(1):145-148.
[18] 葉林,陳政,趙永寧,等. 基于遺傳算法—模糊徑向基神經網絡的光伏發(fā)電功率預測模型[J]. 電力系統(tǒng)自動化,2015,39(16):16-22.