2015 11 Nov 2015 36 6 Journal of Zhengzhou University Engineering Science Vol 36 No 6 1671-6833 2015 06-0056 - 05 C 1 1 2 2 1 450001 2 461000 C FCM FCM MIA MDC MDC MIA I FCM c FCM m FCM C TP18 A doi 10 3969 /j issn 1671-6833 2015 06 011 0 11 MIA MDC 1 FCM m 2 3 1 FCM 1 1 FCM FCM HCM FCM X = x 1 x 2 x n c U = u c m FCM U u j 2 n i i = 1 2 c u K-means 4 5 C FCM 6 j c u GMM 7 i i = 1 SOM 8 SVM 9 i j u 0 1 1 ELM 10 FCM i u > 0 FCM FCM FCM 2015-06 - 10 2015-07 - 19 51307152 1965 E-mail zgs410 @ zzu edu cn
6 C 57 U MIA MDC C min J m U C = c u m d 2 x j c i 2 i = 1 d x j c i = x j - c i 3 MIA MDC n c d c C k k nt k 2 n k k c i = u = c k = 1 u m x j 4 u m { } 2 /-1-1 d ( d ) kj 5 1 2 FCM FCM C U n k 1 step1 c m d CT k C k = n k d 2 CT k nt n k 9 槡 n = 1 ε k = 0 MDC U 0 C 0 d k step2 6 U k i jd k > 0 u k = c = 0u k { ( ) } r = 1 d k 2 /-1-1 d k rj = 1 j k u k = 0 step3 7 C k +1 k +1 ci = u k m x j 6 7 u k m 2 2 2 m step4 C k k +1 - C < ε FCM k = k + 1 step2 m m 1 3 FCM m FCM 1FCM c m 2FCM FCM m m FCM m m m 1 + FCM m 2 Pal 2 FCM 2 1 c c I c FCM CT k k k = 1 2 c MIA MIA c = FCM FCM m 1 槡 c c k = 1 d 2 CT k 8 C k MDC c = mean d CT k CT k 10 ean MDC MIA I I c = MDC c /MIA c 11 MIA c MDC c I c I c c m 1 5 2 5 m FCM
58 2015 12 H m U C FCM I c - 1 > I c c = c + J m U C m 1 step2 H m U C u H m U C = - 1 n c u ln u n i = 1 = 0 u ln u 12 = 0 H m U C m min J m U * C * = arg { 13 s t H m U * C * min J m U C H m U C ( ) μ G = exp - α Jm U * C * max J m U * C * μ C = 1 + β 1 H m U * C * ( ) max H m U * C * 14 15 α > 1 α = 1 5 β β = 10 14 15 = argaxin μ G μ C 16 Fig 1 J m U C H m U C 3 3 1 2 3 FCM FCM m 98 96 I c FCM 96 1 FCM step1 c = 2 I 0 = 0 step2 m = 1 1 J m U * C * H m U * C * step4 m m < 2 5m = m + 0 1 step3 step5 12 c step6 1 2 FCM c I c c step7 2 < c < n I c - 1 > I c - 2 1 FCM The flow chart of adaptive FCM algorithm 96 step3 1 2 FCM c i x i = x i1 x i2 x i96 x' = x max x i1 x i2 x i96 16 0 1
6 C 59 max x i1 x i2 x i96 i 8 FCM x i 3 2 FCM c FCM m 98 FCM 2 FCM FCM 98 c = 8 c = 1 9 4 2 FCM MDC c MIA c I c I c c mfcm J m U * C * H m U * C * J m U C H m U C m FCM 1 M 2005 7-14 2 2 FCM J 2004 37 1 24-28 Fig 2 The classification result of 3 D adaptive FCM algorithm 2012 17-19 4 k - means 2 J 2009 29 4 54-57 5 FCM J 2008 25 6 243-246 FCM c 6 C J 2012 40 m 22 58-62 FCM 7 EM J 2006 11 22 244 FCM - 247 c FCM 8 SOM C - J 8 SOM FCM 2011 23 4 36-39 SOM 9 FCM FCM D 2011 22-27 10 NAGI J YAP K S TIONG S K et al Nontechnical loss
60 2015 detection for metered customers in power utility using support vector machines J IEEE Transactions on Power Delivery 2010 25 2 1162-1171 11 CHICCO G NAPOLI R PETAL P Customer characterization options for improving the tariff offer J IEEE Trans on Power Systems 2003 18 1 381-387 12 BEZDEK J C Pattern recognition with fuzzy objective function algorithms M NewYork Plenum Press 1981 100-136 Power Load Characteristic Classification Technology Research Based on an Optimal Fuzzy C-means Clustering Algorithm ZHAO Guosheng 1 NIU Zhenzhen 1 LIU Yongguang 2 SUN Chaoliang 2 1 School of Electrical Engineering Zhengzhou University Zhengzhou 450001 China 2 Henan Xu Ji Instrument Co Ltd Xuchang 461000 China Abstract In view of the disadvantages of the traditional Fuzzy C-means clustering algorithm the author proposes an adaptive FCM algorithm This algorithm is based on two clustering results evaluation index of within the class distance MIA and between the class distance MDC The ratio of MDC and MIA defined as I is an adaptive function to determine the clustering number c of FCM algorithm At the same time according to the fuzzy decision method we use the objective function and partition entropy of FCM algorithm together to determine the value of optimal fuzzy weighted m This algorithm not only overcomes the FCM algorithm disadvantage of not being able to determine the clustering number automatically and fuzzy weighted index needs to be given by experience but also the clustering result is optimal Finally the correctness and effectiveness of the algorithm were proved through example analysis Key words load clustering FCM load characteristic daily load curve 55 Study on Optimal Allocation Algorithm of the Visible Light Communication LED Array YE Huiying WANG Li LIU Jinliang School of Information Engineering Zhengzhou University Zhengzhou 450001 China Abstract In view of the limitation of the visible light communication light source layout optimization this paper presents a universal optimization layout of light source The simulation results show that using the light source layout of the algorithm not only can make the average area of the room reaches the maximum spectral efficiency and uniform distribution of illumination Illuminance level is between 400 lx and 1500 lx which can meet the demand for lighting Therefore this algorithm can be used as an optimization algorithm for the layout of the universal light source Key words visible light communication LED array layout optimization algorithm universal