王宇譜 呂志平 周海濤 王 寧 翟樹峰
1 信息工程大學(xué)地理空間信息學(xué)院,鄭州市科學(xué)大道62號(hào),450001 2 地理信息工程國家重點(diǎn)實(shí)驗(yàn)室,西安市雁塔路中段1號(hào),710054
?
基于修正鐘差一次差分?jǐn)?shù)據(jù)的衛(wèi)星鐘差預(yù)報(bào)
王宇譜1,2呂志平1周海濤1王 寧1翟樹峰1
1 信息工程大學(xué)地理空間信息學(xué)院,鄭州市科學(xué)大道62號(hào),450001 2 地理信息工程國家重點(diǎn)實(shí)驗(yàn)室,西安市雁塔路中段1號(hào),710054
對(duì)鐘差一次差分預(yù)報(bào)原理進(jìn)行改進(jìn),分析常用的一次多項(xiàng)式模型、二次多項(xiàng)式模型和灰色模型在采用改進(jìn)原理進(jìn)行預(yù)報(bào)時(shí)的相關(guān)特性。結(jié)果表明,對(duì)鐘差一次差分預(yù)報(bào)原理的改進(jìn)是有效的,可以提高常用模型在鐘差短期預(yù)報(bào)中的預(yù)報(bào)精度。
導(dǎo)航衛(wèi)星;鐘差預(yù)報(bào);一次差分;數(shù)據(jù)建模
衛(wèi)星導(dǎo)航系統(tǒng)中星載原子鐘的鐘差預(yù)報(bào)在維持系統(tǒng)的時(shí)間同步、優(yōu)化導(dǎo)航電文中的鐘差參數(shù)等方面具有重要作用[1]。針對(duì)衛(wèi)星鐘差預(yù)報(bào),國內(nèi)外學(xué)者建立了大量模型[2-11]。已有的鐘差預(yù)報(bào)研究主要在于模型本身,本文則通過對(duì)鐘差數(shù)據(jù)進(jìn)行一次差分實(shí)現(xiàn)鐘差建模數(shù)據(jù)與建模策略的改變,研究基于鐘差一次差分?jǐn)?shù)據(jù)的衛(wèi)星鐘差預(yù)報(bào)方法來提高鐘差預(yù)報(bào)的精度。在給出基于鐘差一次差分?jǐn)?shù)據(jù)的預(yù)報(bào)原理及其預(yù)處理方法的基礎(chǔ)上,對(duì)鐘差一次差分預(yù)報(bào)原理進(jìn)行改進(jìn),同時(shí)分析一次多項(xiàng)式(LP)模型、二次多項(xiàng)式(QP)模型和GM(1,1)模型在采用本文原理進(jìn)行預(yù)報(bào)時(shí)的相關(guān)特性。
為了記錄高精度的時(shí)間信息,衛(wèi)星鐘差數(shù)據(jù)的有效位數(shù)通常比較多,數(shù)值相對(duì)較大,使得鐘差數(shù)據(jù)中的異常數(shù)據(jù)點(diǎn)容易被掩蓋。而相鄰歷元間的鐘差數(shù)據(jù)其數(shù)值變化不大,通過歷元間鐘差數(shù)據(jù)的一次差分可以在一定程度上消除原鐘差序列趨勢項(xiàng)的影響,得到一組有效數(shù)字位數(shù)減少、便于進(jìn)行預(yù)處理的數(shù)據(jù)序列。以GPS系統(tǒng)PRN01衛(wèi)星2015-04-02~04-03的15 min采樣間隔的精密鐘差數(shù)據(jù)為例,圖1給出其鐘差數(shù)據(jù),圖2是其對(duì)應(yīng)的一次差分?jǐn)?shù)據(jù)。
圖1 PRN01衛(wèi)星的鐘差數(shù)據(jù)Fig.1 SCB data of satellite PRN01
圖2 PRN01衛(wèi)星的鐘差一次差分?jǐn)?shù)據(jù)Fig.2 Single difference data of SCB of satellite PRN01
對(duì)比圖1和圖2可以看出,原始鐘差數(shù)據(jù)容易掩蓋異常數(shù)據(jù)點(diǎn),而異常數(shù)據(jù)點(diǎn)在其對(duì)應(yīng)的一次差分?jǐn)?shù)據(jù)中表現(xiàn)為峰值點(diǎn),從而使得異常值的探測變得相對(duì)容易。同時(shí),經(jīng)過相鄰歷元間鐘差數(shù)據(jù)的一次差分處理,可以消除原鐘差序列中可能存在的系統(tǒng)誤差。所以,基于鐘差一次差分?jǐn)?shù)據(jù)進(jìn)行鐘差擬合預(yù)報(bào),既便于數(shù)據(jù)的預(yù)處理,也有利于更加準(zhǔn)確地進(jìn)行數(shù)據(jù)擬合。
1.1 基于鐘差一次差分?jǐn)?shù)據(jù)的預(yù)報(bào)原理[1,8-9]
基于鐘差一次差分?jǐn)?shù)據(jù)的預(yù)報(bào)是采用一次差分后的數(shù)據(jù)序列ΔL來對(duì)模型進(jìn)行平差解算。模型確定之后,預(yù)報(bào)n-1個(gè)一次差分建模數(shù)據(jù)對(duì)應(yīng)歷元以后的m(m>1)個(gè)歷元的一次差分序列ΔL1={Δl(j),j=n,n+1, …,n+m-1}。然后將一次差分預(yù)報(bào)序列ΔL1和建模鐘差數(shù)據(jù)的最后一個(gè)鐘差值l(n)對(duì)應(yīng)相加,從而求得所需預(yù)報(bào)歷元j時(shí)刻的鐘差值l(j):
(1)
1.2 基于鐘差一次差分?jǐn)?shù)據(jù)的預(yù)處理方法
針對(duì)鐘差一次差分?jǐn)?shù)據(jù),文獻(xiàn)[7]設(shè)計(jì)了一種基于改進(jìn)中位數(shù)的異常值探測方法:將每個(gè)鐘差一次差分?jǐn)?shù)據(jù)Δl(i)跟一次差分序列的中數(shù)(MED)k及中位數(shù)(MAD)數(shù)倍之和進(jìn)行比較。若鐘差一次差值:
(2)
則認(rèn)為該一次差分?jǐn)?shù)據(jù)是異常值,予以剔除,同時(shí)用內(nèi)插的方法補(bǔ)充該點(diǎn)數(shù)據(jù)。本文采用三次分段樣條方法對(duì)剔除歷元所對(duì)應(yīng)的鐘差一次差值進(jìn)行內(nèi)插補(bǔ)充。式(2)中,k= Median{Δl(i)},MAD=Median{|Δl(i)-k|/0.674 5},n的取值根據(jù)需要確定(本文取n=3)。圖3是對(duì)應(yīng)于圖2的鐘差一次差分?jǐn)?shù)據(jù)預(yù)處理后的數(shù)據(jù),圖4為鐘差一次差分?jǐn)?shù)據(jù)預(yù)處理后對(duì)應(yīng)的鐘差數(shù)據(jù)。從圖3可看出,通過對(duì)異常數(shù)據(jù)點(diǎn)的處理可以得到質(zhì)量更好的鐘差一次差分?jǐn)?shù)據(jù)。
圖3 PRN01衛(wèi)星預(yù)處理后的鐘差一次差分?jǐn)?shù)據(jù)Fig.3 The preprocessed single difference of satellite PRN01
圖4 PRN01衛(wèi)星預(yù)處理后的鐘差數(shù)據(jù)Fig.4 The preprocessed SCB data of satellite PRN01
1.3 基于鐘差一次差分?jǐn)?shù)據(jù)預(yù)報(bào)原理的改進(jìn)
從式(1)可以看出,基于鐘差一次差分?jǐn)?shù)據(jù)的預(yù)報(bào),對(duì)于建模鐘差數(shù)據(jù)中最后一個(gè)數(shù)據(jù)l(n)的依賴程度高,該數(shù)據(jù)的可靠性與基于該預(yù)報(bào)原理的預(yù)報(bào)結(jié)果密切相關(guān)。然而,目前基于鐘差一次差分預(yù)報(bào)原理的鐘差預(yù)報(bào)應(yīng)用中[1,8-9],沒有考慮該問題。
借鑒文獻(xiàn)[11]對(duì)預(yù)報(bào)模型的起點(diǎn)偏差修正所采取的策略,本文在鐘差一次差分預(yù)報(bào)原理的基礎(chǔ)上,通過采用建模數(shù)據(jù)中最后5個(gè)數(shù)據(jù)擬合最后1個(gè)數(shù)據(jù)的方式來提高式(1)中l(wèi)(n)的可靠性,進(jìn)一步完善鐘差一次差分預(yù)報(bào)原理。其中,所采用的擬合公式是鐘差的二次多項(xiàng)式模型,該模型包含了描述鐘差所需的相對(duì)于衛(wèi)星導(dǎo)航系統(tǒng)時(shí)間的偏差、鐘速和鐘漂3種確定性參數(shù)。為了驗(yàn)證所作改進(jìn)的有效性,以GPS系統(tǒng)PRN01衛(wèi)星的鐘差預(yù)報(bào)為例,使用2015-04-06的衛(wèi)星鐘差數(shù)據(jù)進(jìn)行建模,預(yù)報(bào)2015-04-07前6 h的鐘差,分析在給已知建模鐘差數(shù)據(jù)最后一個(gè)值分別加入0.1 ns、0.5 ns、1.0 ns和2.0 ns的模擬粗差時(shí),基于鐘差一次差分預(yù)報(bào)原理改進(jìn)前后的預(yù)報(bào)效果。圖5~8是PRN01衛(wèi)星鐘差在不同模擬粗差條件下的預(yù)報(bào)誤差,基于鐘差一次差分預(yù)報(bào)原理的QP模型記為DQP模型,基于改進(jìn)鐘差一次差分預(yù)報(bào)原理的QP模型記作MDQP模型。表1給出了4種粗差條件下兩種模型預(yù)報(bào)結(jié)果的均方根誤差(RMS,其定義參見式(3))。
圖5 加入0.1 ns粗差時(shí)預(yù)報(bào)結(jié)果的對(duì)比Fig.5 Comparison of prediction results during added outliers with 0.1 ns
由圖表可以看出,MDQP模型的預(yù)報(bào)殘差及RMS值均小于DQP模型,說明改進(jìn)后的鐘差一次差分預(yù)報(bào)原理能夠在一定程度上改善最后一個(gè)建模數(shù)據(jù)不可靠對(duì)預(yù)報(bào)結(jié)果的影響,從而降低對(duì)最后一個(gè)建模鐘差數(shù)據(jù)的依賴程度。
圖6 加入0.5 ns粗差時(shí)預(yù)報(bào)結(jié)果的對(duì)比Fig.6 Comparison of prediction results during added outliers with 0.5 ns
圖7 加入1.0 ns粗差時(shí)預(yù)報(bào)結(jié)果的對(duì)比Fig.7 Comparison of prediction results during added outliers with 1.0 ns
圖8 加入2.0 ns粗差時(shí)預(yù)報(bào)結(jié)果的對(duì)比Fig.8 Comparison of prediction results during added outliers with 2.0 ns
模型粗差值/ns0.10.51.02.0DQP0.1060.4760.9721.971MDQP0.0990.4210.8601.744
為了驗(yàn)證所提的基于鐘差一次差分預(yù)報(bào)原理的有效性,采用IGS提供的GPS系統(tǒng)15 min采樣間隔的精密鐘差數(shù)據(jù)進(jìn)行分析,數(shù)據(jù)采集時(shí)間為2015-04-04~04-18??紤]當(dāng)前在軌運(yùn)行的GNSS星載原子鐘主要是銣原子鐘,而該時(shí)間段內(nèi)GPS星載銣(Rb)鐘包含4種類型:BLOCKⅡA Rb鐘、BLOCK ⅡR Rb鐘、BLOCKⅡR-M Rb鐘和BLOCKⅡF Rb鐘。選取該時(shí)間段內(nèi)數(shù)據(jù)完整的每種類型鐘對(duì)應(yīng)的一顆衛(wèi)星,本文選取PRN01、PRN02、PRN04和PRN12進(jìn)行預(yù)報(bào)實(shí)驗(yàn)。另外,以預(yù)報(bào)時(shí)間段對(duì)應(yīng)的已知精密鐘差值為參考真值,采用均方根誤差(RMS)作為預(yù)報(bào)結(jié)果精度的統(tǒng)計(jì)量進(jìn)行對(duì)比。均方根誤差計(jì)算公式為:
(3)
設(shè)計(jì)兩種預(yù)報(bào)方案,通過對(duì)比常用的LP模型、QP模型和GM(1,1)模型在常規(guī)建模條件下的預(yù)報(bào)結(jié)果和基于改進(jìn)的鐘差一次差分原理的預(yù)報(bào)結(jié)果,分析基于鐘差一次差分預(yù)報(bào)原理的特性。其中,基于改進(jìn)的鐘差一次差分預(yù)報(bào)原理的LP模型和GM(1,1)模型分別記作MDLP模型和MDGM模型,GM(1,1)模型簡記為GM。使用前1 d的衛(wèi)星鐘差數(shù)據(jù)擬合后預(yù)報(bào)接下來6 h的鐘差,連續(xù)預(yù)報(bào)14次。方案1,統(tǒng)計(jì)前4 h的預(yù)報(bào)精度;方案2,統(tǒng)計(jì)前6 h的預(yù)報(bào)精度。
圖9~12分別給出了方案1中4顆衛(wèi)星14次預(yù)報(bào)結(jié)果的RMS值。為了便于分析,表2和表3分別給出了4顆衛(wèi)星在兩種預(yù)報(bào)方案下各模型14次預(yù)報(bào)結(jié)果RMS的平均值及該平均值對(duì)應(yīng)的4顆衛(wèi)星的平均值。
分析圖9~12、表2、表3可知:
1)對(duì)比LP、MDLP模型和QP、MDQP模型結(jié)果可以看出,兩種預(yù)報(bào)方案下每顆衛(wèi)星預(yù)報(bào)結(jié)果的RMS值都是MDLP模型小于LP模型、MDQP模型小于QP模型,說明MDLP模型比LP模型的預(yù)報(bào)精度高、MDQP模型比QP模型的預(yù)報(bào)精度高;對(duì)于4顆衛(wèi)星預(yù)報(bào)精度的平均值而言,兩種方案下MDLP模型較LP模型分別提高了25.5%和11.6%,MDQP模型較QP模型分別提高了17.7%和13.8%;而根據(jù)GM模型和MDGM模型的結(jié)果可知,4顆衛(wèi)星預(yù)報(bào)精度的平均值,在兩種方案下MDGM模型較GM模型分別提高了24.4%和11.0%。因此,在衛(wèi)星鐘差短期預(yù)報(bào)中,采用本文所提的修正鐘差一次差分預(yù)報(bào)原理,可以提高常用的LP模型、QP模型和GM(1,1)模型的預(yù)報(bào)精度。
2)對(duì)比表2和3可以看出,MDLP模型、MDQP模型和MDGM模型在建模數(shù)據(jù)一定的條件下,隨著預(yù)報(bào)時(shí)間的增長,模型預(yù)報(bào)結(jié)果的精度均有所降低。這3種模型中,MDLP模型在兩種預(yù)報(bào)方案下的RMS值都最小,即MDLP模型的預(yù)報(bào)精度優(yōu)于MDQP模型和MDGM模型,說明鐘差一次差分?jǐn)?shù)據(jù)更符合一次多項(xiàng)式模型。此外,根據(jù)各模型對(duì)4種類型衛(wèi)星鐘差預(yù)報(bào)的結(jié)果可知,基于改進(jìn)鐘差一次差分預(yù)報(bào)原理的預(yù)報(bào)結(jié)果與衛(wèi)星鐘的類型有關(guān),特別是對(duì)于BLOCK ⅡF 型銣鐘和ⅡR-M型銣鐘,基于該預(yù)報(bào)原理可以更為顯著地提高常規(guī)鐘差預(yù)報(bào)模型的預(yù)報(bào)精度。
圖9 PRN01衛(wèi)星14次預(yù)報(bào)結(jié)果對(duì)應(yīng)的RMS值Fig.9 RMS values of 14-time prediction results of satellite PRN01
圖10 PRN02衛(wèi)星14次預(yù)報(bào)結(jié)果對(duì)應(yīng)的RMS值Fig.10 RMS values of 14-time prediction results of satellite PRN02
圖11 PRN04衛(wèi)星14次預(yù)報(bào)結(jié)果對(duì)應(yīng)的RMS值Fig.11 RMS values of 14-time prediction results of satellite PRN04
圖12 PRN12衛(wèi)星14次預(yù)報(bào)結(jié)果對(duì)應(yīng)的RMS值Fig.12 RMS values of 14-time prediction results of satellite PRN12
模型統(tǒng)計(jì)值/nsPRN01(ⅡFRb)PRN02(ⅡRRb)PRN04(ⅡARb)PRN12(ⅡR-MRb)平均值LP0.1820.4011.3670.4610.603MDLP0.1160.3591.0410.2790.449QP0.1360.4711.3600.5190.622MDQP0.1330.4121.0600.4440.512GM0.1970.3981.3760.4380.602MDGM0.1110.3631.0670.2780.455
表3 6 h預(yù)報(bào)結(jié)果的統(tǒng)計(jì)值
本文通過對(duì)鐘差數(shù)據(jù)進(jìn)行一次差分實(shí)現(xiàn)鐘差建模數(shù)據(jù)與建模策略的改變,研究基于鐘差一次差分?jǐn)?shù)據(jù)的衛(wèi)星鐘差預(yù)報(bào)方法來提高鐘差預(yù)報(bào)的精度。實(shí)驗(yàn)表明,所提的改進(jìn)的基于鐘差一次差分預(yù)報(bào)原理能夠在一定程度上改善最后一個(gè)建模數(shù)據(jù)含粗差時(shí)對(duì)預(yù)報(bào)結(jié)果的影響;采用基于修正鐘差一次差分的預(yù)報(bào)原理,在鐘差短期預(yù)報(bào)中可以提高常用鐘差預(yù)報(bào)模型的預(yù)報(bào)精度,取得較常用模型更好的預(yù)報(bào)結(jié)果;對(duì)于衛(wèi)星鐘差的一次差分?jǐn)?shù)據(jù)而言,更適合采用一次多項(xiàng)式模型建模。
[1] 王宇譜,呂志平,陳正生,等.衛(wèi)星鐘差預(yù)報(bào)的小波神經(jīng)網(wǎng)絡(luò)算法研究[J].測繪學(xué)報(bào),2013,42(3):20-28(Wang Yupu,Lü Zhiping,Chen Zhengsheng, et al. Research the Algorithm of Wavelet Neural Network to Predict Satellite Clock Bias[J]. Acta Geodaetica et Cartographica Sinica, 2013,42(3):323-330)
[2] 崔先強(qiáng),焦文海.灰色系統(tǒng)模型在衛(wèi)星鐘差預(yù)報(bào)中的應(yīng)用[J].武漢大學(xué)學(xué)報(bào):信息科學(xué)版, 2005,30(5):447-450(Cui Xianqiang,Jiao Wenhai. Grey System Model for the Satellite Clock Error Predicting[J]. Geomatics and Information Science of Wuhan University, 2005, 30(5):447-450)
[3] 鄭作亞,黨亞民,盧秀山,等.附有周期項(xiàng)的預(yù)報(bào)模型及其在GPS衛(wèi)星鐘差預(yù)報(bào)中的應(yīng)用研究[J].天文學(xué)報(bào),2010,51(1):95-102(Zheng Zuoya, Dang Yamin, Lu Xiushan, et al. Prediction Model with Periodic Item and Its Application to the Prediction of GPS Satellite Clock Bias[J]. Acta Astronomica Sincia, 2010, 51(1):95-102)
[4] Heo Y J, Cho J, Heo M B. Improving Prediction Accuracy of GPS Satellite Clocks with Periodic Variation Behavior[J]. Measurement Science and Technology, 2010, 21(7): 3 001-3 008
[5] Senior K L, Ray J R, Beard R L. Characterization of Periodic Variations in the GPS Satellite Clocks[J]. GPS Solutions, 2008, 12(3):211-225
[6] Davis J, Bhattarai S, Ziebart M. Development of a Kalman Filter Based GPS Satellite Clock Time-offset Prediction Algorithm [C]. European Frequency and Time Forum (EFTF), Gothenburg, Sweden, 2012
[7] 王宇譜,呂志平,宮曉春,等.一種新的鐘差預(yù)處理方法及在WNN鐘差中長期預(yù)報(bào)中的應(yīng)用[J].武漢大學(xué)學(xué)報(bào):信息科學(xué)版, 2016,41(3): 373-379(Wang Yupu,Lü Zhiping,Gong Xiaochun, et al. A New Data Preprocessing Method for Satellite Clock Bias and Its application in WNN to Predict Medium- and Long-term Clock Bias[J]. Geomatics and Information Science of Wuhan University, 2016,41(3): 373-379)
[8] 雷雨,趙丹寧.基于經(jīng)驗(yàn)?zāi)J椒纸夂妥钚《酥С窒蛄繖C(jī)的衛(wèi)星鐘差預(yù)報(bào)[J].天文學(xué)報(bào),2014,55(3):216-227(Lei Yu, Zhao Danning. The Satellite Clock Bias Forecast Based on Empirical Mode Decomposition and Least Squares Support Vector Machines[J]. Acta Astronomica Sinica, 2014,55(3):216-227)
[9] 梁月吉,任超,楊秀發(fā),等.基于一次差的灰色模型在衛(wèi)星鐘差預(yù)報(bào)中的應(yīng)用[J].天文學(xué)報(bào),2015, 56(3): 264-277(Liang Yueji, Ren Chao, Yang Xiufa, et al. Grey Model Based on First Difference in the Application of the Satellite Clock Bias Prediction[J]. Acta Astronomica Sinica, 2015, 56(3): 264-277)
[10]郭海榮.導(dǎo)航衛(wèi)星原子鐘時(shí)頻特性分析理論與方法研究[D].鄭州: 信息工程大學(xué), 2006(Guo Hairong. Study on the Analysis Theories and Algorithms of the Time and Frequency Characterization for Atomic Clocks of Navigation Satellite[D]. Zhengzhou: Information Engineering University, 2006)
[11]黃觀文. GNSS星載原子鐘質(zhì)量評(píng)價(jià)及精密鐘差算法研究[D].西安: 長安大學(xué),2012(Huang Guanwen. Research on Algorithms of Precise Clock Offset and Quality Evaluation of GNSS Satellite Clock[D]. Xi’an: Chang ’an University, 2012)
About the first author:WANG Yupu, PhD candidate, majors in theory and method of satellite clock data processing, E-mail: 987834660@qq.com.
Satellite Clock Bias Prediction Based on Modified Single Difference Data of Clock Bias
WANGYupu1,2LüZhiping1ZHOUHaitao1WANGNing1ZHAIShufeng1
1 School of Surveying and Mapping, Information Engineering University, 62 Kexue Road, Zhengzhou 450001, China 2 State Key Laboratory of Geo-Information Engineering, 1 Mid-Yanta Road, Xi’an 710054, China
An improved strategy for the prediction principle based on single difference values of satellite clock bias (SCB) is proposed. This paper analyzes prediction characteristics of the frequently-used linear polynomial model, quadratic polynomial model and grey model when they use the proposed prediction principle. The simulation results show that the proposed improvement for the prediction principle is effective and the prediction precision of the frequently-used models can get better prediction results based on the proposed prediction principle in the short-term SCB prediction.
navigation satellite; clock bias prediction; single difference; data modeling
National Natural Science Foundation of China, No. 41674019; National Key Research and Development Program of China, No.2016YFB0501701; Open Fund of State Key Laboratory of Geo-Information Engineering, No. SKLGIE2015-M-1-6.
2015-12-23
項(xiàng)目來源:國家自然科學(xué)基金(41674019);國家重點(diǎn)研發(fā)計(jì)劃(2016YFB0501701);地理信息工程國家重點(diǎn)實(shí)驗(yàn)室開放基金(SKLGIE2015-M-1-6)。
王宇譜,博士生,主要從事衛(wèi)星鐘數(shù)據(jù)的處理理論與方法研究,E-mail: 987834660@qq.com。
10.14075/j.jgg.2016.12.009
1671-5942(2016)012-1073-05
P228
A