国产日韩欧美一区二区三区三州_亚洲少妇熟女av_久久久久亚洲av国产精品_波多野结衣网站一区二区_亚洲欧美色片在线91_国产亚洲精品精品国产优播av_日本一区二区三区波多野结衣 _久久国产av不卡

?

基于混沌反向?qū)W習(xí)和水波算法改進(jìn)的白鯨優(yōu)化算法

2024-05-24 02:14:02王亞輝張虎晨王學(xué)兵胡繼明李婭

王亞輝 張虎晨 王學(xué)兵 胡繼明 李婭

摘 要:針對(duì)原始的白鯨算法(beluga whale optimization,BWO)在某些情況下,中后期的探索和開(kāi)發(fā)能力不足、多樣性和求解精度降低、容易陷入局部最優(yōu)等問(wèn)題,提出一種基于混沌反向?qū)W習(xí)和水波算法改進(jìn)的白鯨優(yōu)化算法(TWBWO),進(jìn)一步提高白鯨算法的計(jì)算精度和收斂速度,增強(qiáng)全局搜索和跳出局部最優(yōu)能力。結(jié)合混沌映射和反向?qū)W習(xí)策略提高種群的質(zhì)量和多樣性,加快收斂速度。引入水波算法(water wave optimization,WWO)的折射操作,避免尋優(yōu)時(shí)輕易陷入局部最優(yōu),提高計(jì)算精度。實(shí)驗(yàn)結(jié)果表明,TWBWO算法較之原始算法和其他經(jīng)典算法在收斂速度和求解精度以及穩(wěn)定性方面更為優(yōu)秀,性能和尋優(yōu)能力更強(qiáng)。

關(guān)鍵詞:白鯨優(yōu)化算法; 水波算法; 混沌映射; 反向?qū)W習(xí); 算法改進(jìn)

中圖分類號(hào):TP301?? 文獻(xiàn)標(biāo)志碼:A

文章編號(hào):1001-3695(2024)03-013-0729-07

doi:10.19734/j.issn.1001-3695.2023.07.0332

Improved beluga whale optimization algorithm based onchaotic inverse learning and water wave algorithm

Wang Yahui, Zhang Huchen, Wang Xuebing, Hu Jiming, Li Ya

(College of Mechanical, North China University of Water Resources & Electric Power, Zhengzhou 450045, China)

Abstract:To address the problems of the original BWO algorithm,such as insufficient exploration and exploitation ability in the middle and late stages in some cases, reduce the diversity and solution accuracy, easy to fall into local optimality,this paper proposed a white whale optimization algorithm (TWBWO) based on chaotic backward learning and water wave algorithm improvement. Further it improved the computational accuracy and convergence speed of Moby Dick algorithm, enhanced the ability of global search and jumping out of local optimum. Combining chaotic mapping and backward learning strategies,it improved the quality and diversity of populations and speeded up the convergence rate. It introduced the refraction operation of the WWO to avoid the algorithm from repeatedly falling into local optima and improve the computational accuracy. The experimental results show that the TWBWO algorithm is superior to the original algorithm and other classical algorithms in terms of convergence speed and solution accuracy as well as stability, with better performance and better finding ability.

Key words:BWO; WWO; chaotic mapping; reverse learning; algorithm improvement

0 引言

尋得問(wèn)題的最佳處理方案一直是各學(xué)科的焦點(diǎn),受自然界生物的啟發(fā),學(xué)者們提出了很多群智能優(yōu)化算法來(lái)解決最優(yōu)問(wèn)題[1]。白鯨算法(BWO)是由Zhong等人[2]觀察白鯨群體的生活行為后于2022年提出的一種元啟發(fā)式優(yōu)化算法,該算法具有結(jié)構(gòu)設(shè)計(jì)簡(jiǎn)單、收斂速度快、求解精度高等優(yōu)點(diǎn)。黃欣等人[3]使用白鯨算法優(yōu)化密度聚類算法(density-based spatial clustering of applications with noise,DBSCAN)實(shí)現(xiàn)全局參數(shù)的自適應(yīng)選取,提升了效率和聚類結(jié)果的可信度,并應(yīng)用于變電站數(shù)據(jù)流異常檢測(cè)。Houssein等人[4]將動(dòng)態(tài)候選解(dynamic candidate solutions,DCS)和K-最近鄰(K-nearest neighbor,KNN)分類器結(jié)合,通過(guò)給潛在候選者一個(gè)機(jī)會(huì),提高了選擇解的多樣性和一致性。蔡海良等人[5]利用白鯨算法優(yōu)化極限學(xué)習(xí)機(jī)(extreme learning machine,ELM)初始參數(shù),用于電力光纜故障診斷及定位。

針對(duì)傳統(tǒng)白鯨算法在一些情況下多樣性和求解精度降低、容易陷入局部最優(yōu)等問(wèn)題,本文提出一種基于混沌反向?qū)W習(xí)和水波算法(WWO)改進(jìn)的白鯨算法(TWBWO)?;煦缬成浜头聪?qū)W習(xí)用于前期的種群初始化來(lái)提升種群的質(zhì)量和多樣性,加快收斂速度。引入水波算法的折射理念,折射當(dāng)前迭代尋到的最優(yōu)解,提高計(jì)算精度,避免陷入局部最優(yōu)。通過(guò)多組測(cè)試函數(shù)、多維度、多算法進(jìn)行對(duì)比實(shí)驗(yàn)。結(jié)果表明,改進(jìn)后的算法較原始算法性能有明顯提升。

1 原始白鯨優(yōu)化算法(BWO)

BWO算法是通過(guò)觀察白鯨生活行為提出的元啟發(fā)式算法,如圖1所示。

1.1 初始化

基于白鯨種群的機(jī)制,白鯨被視為搜索代理,每條白鯨都代表一個(gè)候選方案,在優(yōu)化過(guò)程中不斷更新,其模型被建立為

2.2 水波算法(WWO)

Zheng受淺水波理論的啟發(fā),模擬水波運(yùn)動(dòng)中的傳播、折射、碎浪運(yùn)動(dòng),于2015年提出了水波優(yōu)化算法(WWO)[14]。該算法中水波的適應(yīng)度值與海底深度和波長(zhǎng)成反比,距離靜止水位越近,適應(yīng)度值越高,波長(zhǎng)越短[15]。通過(guò)波長(zhǎng)控制水波的搜索范圍,適應(yīng)度低的水波進(jìn)行大范圍的搜索,最大可能尋找更優(yōu)解,適應(yīng)度高的水波進(jìn)行小范圍局部搜索,保證解的質(zhì)量,如圖4[16]所示。

每一次迭代,種群中首先會(huì)進(jìn)行傳播操作,水波個(gè)體x通過(guò)在每一維d上增加隨機(jī)的位移來(lái)產(chǎn)生新的個(gè)體解[17]。每次的傳播操作更新波長(zhǎng)之后會(huì)衰減水波個(gè)體的波高,當(dāng)波高逐漸降低為0時(shí),算法進(jìn)入折射操作,使其向當(dāng)前最優(yōu)個(gè)體學(xué)習(xí),避免算法停滯[18]。在搜索到新的最優(yōu)水波后進(jìn)行碎浪操作,并將其分解,進(jìn)行更細(xì)致的局部搜索,保證解的優(yōu)越性。

為了提升BWO算法跳出局部最優(yōu)的能力,提高收斂速度與計(jì)算精度,將WWO與BWO算法結(jié)合。在BWO算法尋到當(dāng)前全局最優(yōu)解后,引入折射操作,對(duì)最優(yōu)解進(jìn)行折射學(xué)習(xí),對(duì)比新個(gè)體與當(dāng)前最優(yōu)個(gè)體適應(yīng)度值,選擇優(yōu)值替代進(jìn)入下一次迭代。改進(jìn)后公式如下:

Xd=N(Xbest+Xm2,|Xbest-Xm|2)(14)

Xm=mean(Xbest+Xworst)(15)

其中:Xbest、Xworst分別為當(dāng)前種群中最優(yōu)和最差個(gè)體;Xm是兩者的平均位置;N(μ,σ)是均值為μ、標(biāo)準(zhǔn)差為σ的高斯隨機(jī)數(shù)。

圖5為水波算法優(yōu)化后的BWO算法與原始BWO算法在同一測(cè)試函數(shù)上的迭代曲線對(duì)比??梢钥闯觯Y(jié)合水波算法后,白鯨算法跳出局部最優(yōu)的能力明顯提升,加快了收斂速度,提升了計(jì)算精度。

2.3 TWBWO算法

混沌反向?qū)W習(xí)和水波算法都能對(duì)白鯨算法進(jìn)行有效的優(yōu)化,在一定程度上提升算法性能,但兩者各自所優(yōu)化BWO算法的方式并不沖突且互不影響?;煦绶聪?qū)W習(xí)策略優(yōu)化BWO算法的初始化階段,提升種群質(zhì)量。水波算法影響的是勘探、開(kāi)發(fā)、鯨落階段的當(dāng)前最優(yōu)值。因此,將兩者互補(bǔ)結(jié)合,進(jìn)一步提升BWO算法的性能。在同一測(cè)試函數(shù)上的實(shí)驗(yàn)對(duì)比結(jié)果如圖6所示。可以看出,TWBWO算法充分繼承了前兩者的優(yōu)點(diǎn),進(jìn)一步提升了BWO算法的性能。

TWBWO算法流程如圖7所示。TWBWO算法實(shí)現(xiàn)步驟如下:

a)對(duì)參數(shù)P、dim、lb、ub、T進(jìn)行初始化,P為種群中個(gè)體數(shù)量,dim為個(gè)體維度,lb、ub為解的下限和上限,T為最大迭代次數(shù)。

b)使用混沌反向?qū)W習(xí)策略進(jìn)行初始化,獲得新的初始種群,檢驗(yàn)是否超出搜索邊界,計(jì)算適應(yīng)度值并排序,得到當(dāng)前最優(yōu)個(gè)體解和最差個(gè)體解。

c)根據(jù)式(3)(10)計(jì)算平衡因子Bf和鯨落概率Wf。

d)如果Bf<0.5,進(jìn)入開(kāi)發(fā)階段,按照式(5)更新個(gè)體位置;如果Bf>0.5,進(jìn)入勘探階段,按照式(4)更新個(gè)體位置。

e)在步驟c)d)結(jié)束后,對(duì)比平衡因子Bf和鯨落概率Wf。若BfWf,則直接檢測(cè)是否超出邊界。

f)在所有個(gè)體經(jīng)歷過(guò)一次迭代后,計(jì)算適應(yīng)度值,找出新的當(dāng)前最優(yōu)解和最差解,根據(jù)式(14)(15)對(duì)最優(yōu)解進(jìn)行折射操作,計(jì)算適應(yīng)度值并對(duì)比保存最優(yōu)解。

g)判斷是否滿足最大迭代次數(shù)或停止條件,不滿足則返回步驟c),直至滿足后輸出最終結(jié)果。

3 實(shí)驗(yàn)與分析

3.1 實(shí)驗(yàn)設(shè)計(jì)

選擇十個(gè)基準(zhǔn)測(cè)試函數(shù)進(jìn)行實(shí)驗(yàn),測(cè)試改進(jìn)算法的性能。同時(shí),為對(duì)比TWBWO算法的優(yōu)越性,將其與原始白鯨算法(BWO)、鯨魚算法(whale optimization algorithm,WOA)[19]、灰狼算法(grey wolf optimizer,GWO)[20]、混合蛙跳算法(grey wolf optimizer,SFLA)[21]、果蠅算法(fruit fly optimization algorithm,F(xiàn)OA)[22]進(jìn)行對(duì)比實(shí)驗(yàn)。實(shí)驗(yàn)中種群規(guī)模為50,最多迭代100次,分別進(jìn)行30次獨(dú)立實(shí)驗(yàn)。測(cè)試函數(shù)如表1所示。

表1給出了測(cè)試函數(shù)的表達(dá)式、維度、范圍以及理論最優(yōu)值,其中F1~F4為單模態(tài)測(cè)試函數(shù),用于測(cè)試算法是否能快速搜尋到最優(yōu)解;F5~F10為多模態(tài)測(cè)試函數(shù),用于測(cè)試算法能否有效跳出局部最優(yōu)[23]。為了更好地體現(xiàn)算法在不同維度上的性能,將測(cè)試函數(shù)的維度分別設(shè)置為30、100、500維。

3.2 實(shí)驗(yàn)結(jié)果

表2~4分別列出了TWBWO、BWO、WOA、GWO、FOA和SFLA算法在十個(gè)測(cè)試函數(shù)的30、100、500維度上獨(dú)立運(yùn)行30次,每次迭代100步的實(shí)驗(yàn)結(jié)果。實(shí)驗(yàn)結(jié)果統(tǒng)計(jì)了每個(gè)算法運(yùn)行30次的最優(yōu)值(best)、平均值(mean)、標(biāo)準(zhǔn)差(std),在尋找最優(yōu)解的問(wèn)題中,平均值越小表明算法尋優(yōu)精度越高,而標(biāo)準(zhǔn)差越小則證明算法越穩(wěn)定[24]。

由表中結(jié)果可知,通過(guò)平均值的比較,無(wú)論是在4個(gè)單模態(tài)測(cè)試函數(shù),還是6個(gè)多模態(tài)測(cè)試函數(shù)中,TWBWO算法在尋優(yōu)精度上都表現(xiàn)出了更好的性能,證明結(jié)合水波算法的折射操作可以有效提升白鯨算法的局部搜索能力。通過(guò)標(biāo)準(zhǔn)差的比較,驗(yàn)證了TWBWO算法的穩(wěn)定性更好。而在面對(duì)高維問(wèn)題時(shí),求解過(guò)程更加復(fù)雜,算法尋找最優(yōu)解的難度不斷提升,TWBWO算法仍保持較高的尋優(yōu)精度和魯棒性,較之其他算法更加穩(wěn)定。

為了驗(yàn)證TWBWO算法在收斂速度方面的性能,針對(duì)本文使用的十個(gè)測(cè)試函數(shù),選擇尋優(yōu)維度為30,以迭代次數(shù)為x軸,適應(yīng)度值為y軸,繪制TWBWO、BWO、WOA、GWO、FOA、SFLA六個(gè)算法的迭代曲線進(jìn)行對(duì)比[25],如圖8所示。

在十個(gè)測(cè)試函數(shù)中,TWBWO算法的收斂速度明顯優(yōu)于其他算法。在單模態(tài)測(cè)試函數(shù)中,TWBWO算法在函數(shù)F1、F2明顯收斂更快,雖無(wú)法收斂,但得到的最終結(jié)果在10-40上下,已無(wú)限接近最優(yōu)解。函數(shù)F3未在100次迭代后收斂,但相較其他算法,在都無(wú)法收斂的情況下,前期搜索效率更高、速度更快且更接近最優(yōu)解。在多模態(tài)測(cè)試函數(shù)中,TWBWO算法只需20次迭代便可在函數(shù)F5收斂得到最優(yōu)解,且精度最好,更是只需38次迭代和42次迭代便得到了函數(shù)F6和F8的理論最優(yōu)解,在函數(shù)F7、F9、F10的收斂速度也都有明顯提高。

通過(guò)混沌反向?qū)W習(xí)策略加強(qiáng)了初始種群的質(zhì)量和多樣性,使得算法一開(kāi)始便能找到優(yōu)秀的初始值,在迭代初期就有較高的精度。折射操作的存在也能保證算法不會(huì)重復(fù)陷入局部最優(yōu),加快收斂速度。這使得TWBWO算法在收斂速度和尋優(yōu)精度上都具有明顯的優(yōu)越性。

穩(wěn)定性是評(píng)價(jià)算法優(yōu)劣的一個(gè)重要因素,為了驗(yàn)證TWBWO算法的穩(wěn)定性,對(duì)TWBWO算法和五個(gè)對(duì)比算法在F3測(cè)試函數(shù)上進(jìn)行30次獨(dú)立實(shí)驗(yàn),收集實(shí)驗(yàn)結(jié)果并繪制箱型圖,結(jié)果如圖9所示。從箱型圖可以看出,TWBWO算法30次獨(dú)立實(shí)驗(yàn)所得到的最優(yōu)值分布基本在一條直線上,相比于BWO算法分布波動(dòng)更小、更均勻,對(duì)比另外四個(gè)算法更是有著巨大的優(yōu)勢(shì),證明TWBWO算法具有更為優(yōu)越的穩(wěn)定性。

4 結(jié)束語(yǔ)

針對(duì)原始白鯨算法在某些情況下,中后期的探索和開(kāi)發(fā)能力不足、多樣性和求解精度降低、容易陷入局部最優(yōu)等問(wèn)題,提出一種基于混沌反向?qū)W習(xí)和水波算法改進(jìn)的白鯨優(yōu)化算法(TWBWO)來(lái)提高計(jì)算精度和收斂速度,增強(qiáng)全局搜索和跳出局部最優(yōu)的能力。算法初期通過(guò)混沌映射和反向?qū)W習(xí)策略初始化種群,提高種群的質(zhì)量和多樣性,提高初始解的質(zhì)量。引入水波算法(WWO)的折射操作對(duì)每次迭代得到的最優(yōu)解進(jìn)行折射,增強(qiáng)算法跳出局部最優(yōu)的能力,提高算法的計(jì)算精度,進(jìn)一步提升算法的尋優(yōu)性能。

實(shí)驗(yàn)結(jié)果表明,即使在高維情況下,TWBWO算法在收斂速度和求解精度以及穩(wěn)定性方面都更為優(yōu)秀,尋優(yōu)性能更強(qiáng)。相較于提高初始種群質(zhì)量和折射當(dāng)前最優(yōu)解來(lái)提高算法性能,如何改進(jìn)算法的勘探階段和開(kāi)發(fā)階段進(jìn)一步提升算法性能,是下一步工作研究的重點(diǎn)。

圖9 F3最優(yōu)解分布箱型圖

Fig.9 Box plot of F3 optimal solution distribution

參考文獻(xiàn):

[1]Ouyang Weimin, Ramachandran V. Classroom education effect evaluation model based on MFO intelligent optimization algorithm[J]. Journal of Intelligent & Fuzzy Systems, 2021,40(4) : 6791-6802.

[2]Zhong Changting, Li Gang, Meng Zeng. Beluga whale optimization: a novel nature-inspired metaheuristic algorithm[J]. Knowledge-Based Systems, 2022,251: article No.109215.

[3]黃欣, 趙敏彤, 郇嘉嘉, 等. 基于BWO-DBSCAN和CSA-OCRKELM的變電站數(shù)據(jù)流異常檢測(cè)方法[J]. 廣東電力, 2023,36(5): 39-48. (Huang Xin, Zhao Mintong, Jun Jiajia, et al. A substation data flow anomaly detection method based on BWO-DBSCAN and CSA-OCRKELM[J]. Guangdong Electric Power, 2023,36(5): 39-48.)

[4]Houssein E H. Sayed A. Dynamic candidate solution boosted beluga whale optimization algorithm for biomedical classification[J]. Mathematics, 2023,11(3): article No.707.

[5]蔡海良, 胡凱, 李軍, 等. 基于BWO-ELM算法與VR-GIS技術(shù)的電力光纜故障診斷及定位研究[J]. 計(jì)算機(jī)測(cè)量與制, 2022, 30(12): 98-104,111. (Cai Hailiang, Hu Kai, Li Jun, et al. Research on power fiber optic cable fault diagnosis and localization based on BWO-ELM algorithm and VR-GIS technology[J]. Computer Measurement and System, 2022,30(12): 98-104,111.

[6]Ma Denglong, Tan Wei, Wang Qingsheng, et al. Application and improvement of swarm intelligence optimization algorithm in gas emission source identification in atmosphere[J]. Journal of Loss Prevention in the Process Industries, 2018,56: 262-271.

[7]Zhang Jinghua, Dong Ze. A general intelligent optimization algorithm combination framework with application in economic load dispatch problems[J]. Energies, 2019,12(11): article No.2175.

[8]Gurses D, Pholdee N, Bureerat S, et al. A novel hybrid water wave optimization algorithm for solving complex constrained engineering problems[J]. Materials Testing, 2021,63(6):560-564.

[9]Yin Qiang, Cao Ben, Li Xue,et al. An intelligent optimization algorithm for constructing a DNA storage code: NOL-HHO[J]. International Journal of Molecular Sciences, 2020,21(6) : article No.2191.

[10]Cui Fan, Ni Jianyu, Du Yunfei, et al. Soil water content estimation using ground penetrating radar data via group intelligence optimization algorithms: an application in the northern shaanxi coal mining area[J]. Energy Exploration & Exploitation, 2021,39(1): 318-335.

[11]Nie Xin, Luo Jian. The hybrid intelligent optimization algorithm and multi-objective optimization based on big data[J]. Journal of Phy-sics: Conference Series, 2021,1757(1) : article No.012132.

[12]Hu Shanshan, Liu Hui, Feng Yufei, et al. Tool wear prediction in glass fiber reinforced polymer small-hole drilling based on an improved circle chaotic mapping grey wolf algorithm for BP neural network[J]. Applied Sciences, 2023,13(5): article No.2811.

[13]Huang Yujie, Li Yibing, Zhang Zitang, et al. A novel path planning approach for AUV based on improved whale optimization algorithm using segment learning and adaptive operator selection[J]. Ocean Engineering, 2023,280: article No.114591.

[14]付接遞. 鯨魚優(yōu)化算法改進(jìn)與應(yīng)用研究[D]. 銀川:寧夏大學(xué), 2022. (Fu Jiedi. Research on the improvement and application of whale optimization algorithm[D]. Yinchuan:Ningxia University, 2022.)

[15]Sun Yulong, Li Hongjuan, Shabaz M,et al. Research on building truss design based on particle swarm intelligence optimization algorithm[J]. International Journal of System Assurance Enginee-ring and Management, 2021,13: 38-48.

[16]Azam A, Bardhan A, Kaloop M R, et al. Modeling resilient modulus of subgrade soils using LSSVM optimized with swarm intelligence algorithms[J]. Scientific Reports, 2022,12: article No.14454.

[17]Ozbay F A. A modified seahorse optimization algorithm based on chao-tic maps for solving global optimization and engineering problems[J]. Engineering Science and Technology, an International Journal, 2023,41: 1-26.

[18]Zhang Danfeng, Han Congai, Zhang Haiyan, et al. The simulation design of microwave absorption performance for the multi-layered carbon-based nanocomposites using intelligent optimization algorithm[J]. Nanomaterials, 2021,11(8): article No.1951.

[19]Li Yuzhou, Sun Chuanxia, Hu Yinglei. Whale optimization algorithm-based deep learning model for driver identification in intelligent transport systems[J]. Computers, Materials & Continua, 2023,75(2):3497-3515.

[20]Lambrinidis G, Tsantili-Kakoulidou A. Multi-objective optimization methods in novel drug design[J]. Expert Opinion on Drug Disco-very, 2021,16(6): 647-658.

[21]Zhang Meilun, Zhang Jin, Xu Maoda, et al. Improved SFLA algorithm with hybrid SA strategy for multi-objective reactive power planning optimization of distribution network[J]. Journal of Physics: Conference Series, 2021,1865(4) : article No.042056.

[22]Yonghua Bai, Minzhou Luo, Pang Fenglin. An algorithm for solving robot inverse kinematics based on FOA optimized BP neural network[J]. Applied Sciences, 2021,11(15): article No.7129.

[23]Jiang Wenjun, Zhang Ning. Research on characteristics of paper-plastic composite film based on intelligent optimization algorithm[J]. Personal and Ubiquitous Computing, 2021,25(S1): article No.27.

[24]Zhang Junrong, Tang Huiming, Tannant D D, et al. A novel model for landslide displacement prediction based on EDR selection and multi-swarm intelligence optimization algorithm[J]. Sensors, 2021,21(24): article No.8352.

[25]Shokouhandeh H, Kamarposhti M, Holderbaum W, et al. Optimal operation of distributed generations considering demand response in a microgrid using GWO algorithm[J]. Computer Systems Science and Engineering, 2023,47(1):809-822.

屏南县| 新源县| 文昌市| 汉川市| 平邑县| 云霄县| 黔东| 高邮市| 平安县| 伊金霍洛旗| 类乌齐县| 张家口市| 康乐县| 安平县| 专栏| 大化| 洛川县| 城口县| 布尔津县| 苗栗市| 汉阴县| 湟源县| 蓝山县| 万年县| 文水县| 育儿| 西和县| 乐昌市| 南和县| 攀枝花市| 彭泽县| 育儿| 福州市| 札达县| 保定市| 石城县| 阆中市| 元氏县| 珲春市| 阿拉善右旗| 宁海县|