Inter national J our nal of Electrical and Computer Engineering (IJECE) V ol. 6, No. 6, December 2016, pp. 3276 3282 ISSN: 2088-8708 3276 Con v er gence Ev aluation of a Load Flo w Method based on Cespedes’ A ppr oach to Distrib ution System Analysis Diego Issicaba 1 and J or ge Coelho 2 1 Department of Electrical Engineering, Federal Uni v ersity of T echnology - P arana (UTFPR), Curitiba-PR, Brazil 2 Department of Electrical Engineering, Federal Uni v ersity of Santa Catarina (UFSC), Florianopolis-S C, Brazil Article Inf o Article history: Recei v ed Aug 9, 2016 Re vised Sep 20, 2016 Accepted Oct 7, 2016 K eyw ord: Po wer engineering Po wer distrib ution systems Load flo w analysis Con v er gence ABSTRA CT This paper e v aluates the con v er gence of a load flo w met hod based on Cespedes’ for - mulation to distrib ution system steady-state analysis. The method is desc ribed and the closed-form of its con v er gence rate is deduced. Furthermore, con v er gence dependence of loading and the consequences of choosing particular initial estimates are v erified mathematically . All mathematical res ults ha v e been tested in numerical simulations, some of them presented in the paper . Copyright c 2016 Institute of Advanced Engineering and Science . All rights r eserved. Corresponding A uthor: Die go Issicaba Department of Electrical Engineering, Federal Uni v ersity of T echnology - P arana (UTFPR) A v . Sete de Setembro, 3165, Sector D, Rebouc ¸ as, 80230-910 Curitiba-PR, Brazil +55 41 3310-4626 issicaba@utfpr .edu.br 1. INTR ODUCTION Load flo w methods are fundamental tools to po wer distrib ution system analysis [1, 2]. These methods allo w computing steady-state v oltages at netw ork nodes as well as the amount of po wer flo wing through po wer system de vices. Ne v ertheless, po wer system literature lacks on formal analysis and comparisons among con- v er gence properties of load flo w algorithms. Consistent e xceptions can be foun d in [3, 4, 5, 6, 7]. P articularly , in [6] the con v er gence of a forw ard-backw ard sweep method is e v aluated and its dependence on system loading v erified. In [7], this same method is formally assessed using fix ed-point concepts and the contraction mapping theorem. In this conte xt, this paper introduces the con v er gence analysis of a load flo w method based on R. G. Cespedes’ approach to distrib ution system analysis [8]. F or t his accomplishment, section 2 introduces the proposed method focusing on algorithm procedures and main equations. Section 3 presents the deduction of the con v er gence rate of the algorithm and a mathematical re gion where algorithm iterates are confined, as long as initial estimates are chosen properly . In section 4, numerical results are sho wn to illustrate the v alidity of the mathematical de v elopments. At last, section 5 outlines conclusions and final remarks. 2. A LO AD METHOD INSPIRED ON CESPEDES’ FORMULA TION Consider the radial feeder schematic sho wn in Fig. 1. In the schematic, z i denotes a series line impedance, E i = e i + j f i represents comple x node v oltages, and S L i denotes comple x loads, all refereed to a general node i . The inde x u i stands for the node upstream node i. The substation b us is named the 0 (zero) node with comple x v oltage denoted by E 0 . Furthermore, the inde x i points out to both a node and its upstream line. The comple x po wer flo w do wnstream line i can be computed by summing the loads and l osses do wn- J ournal Homepage: http://iaesjournal.com/online/inde x.php/IJECE DOI:  10.11591/ijece.v6i6.12134 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 3277 q q ? S L i q ? S L i +1 q ? S L n 1 q ? S L n E u i E i E i +1 E N 1 E N z i z i +1 z n S i - S i |{z} branch n Figure 1. Radial distrib ution netw ork schematic (adapted from [1]). stream the line as follo ws S i = S L i + X r 2 i S L r + X r 2 i z r S r E r 2 (1) where i represents either the set of nodes do wnstream node i or do wnstream line i , depending on the v ariables in v olv ed. As consequence, S i can be e xpressed as function of v ariables associated to nodes immediately do wnstream line i , as follo ws S i = S L i + X d 2D i S d + X d 2D i z d S d E d 2 (2) where D i denotes either the set of nodes immediately do wnstream node i or immediately do wnstream line i . Furthermore, the dif ference between node v oltages in adjacent nodes can be e xpressed as E i = E u i z i S i E i (3) Separating the real and imaginary parts of (2) and the v oltage magnitudes in (3), the recursi v e equa- tions utilized in the load flo w method proposed by R. G. Cespedes [8] can be deduced. Con v ersely , by applying (2) and (3) without further deductions, a similar b ut also ef fecti v e load flo w method can be designed. Starting from initial comple x v oltage estimates, the tw o steps belo w can be continually repeated until the con v er gence of comple x v oltages is reached. 1. In a backw ard process, the comple x po wer flo w at each node is calculated using (2), starting at end-nodes and stopping at the first node immediately do wnstream from the substation node. 2. In a forw ard process, comple x v oltage are updated, a w ay from the substation node, using (3). The comple x v oltage at the substation b us is assumed constant during the procedures. The con v er gence criterion refers to the maximum absolute mismatch between subsequent comple x v oltage iterates. 3. CONVERGENCE ASSESSMENT This section addresses the con v er ge assessment of the algorithm, focusing on aspects related to the implications of choosing an initial estimate and the deduction of the con v er gence rate. 3.1. Initial Estimate The update rule of the algorithm can be written recursi v ely for iteration k as   ( k ) i =   ( k ) u i z i E ( k ) i S i;c + L ( k ) i;ac (4) where E ( k ) i denotes the comple x v oltage at node i and iteration k ,   ( k ) i is the comple x v oltage at node i and iteration k + 1 , S i;c denotes the sum of all comple x loads do wnstream node i (including the one at node i ) and L ( k ) i;ac represents the sum of all electrical losses do wnstream node i and iteration k . Con ver g ence Evaluation of a Load Flow Method based on Cespedes’ Appr oac h ... (Die go Issicaba) Evaluation Warning : The document was created with Spire.PDF for Python.
3278 ISSN: 2088-8708 Let us define ~ i as the set of lines in the path between node 0 and i . By (4), we ha v e   ( k ) 1 = E 0 z 1 E ( k ) 1 S 1 ;c + L ( k ) 1 ;ac . . . =   ( k ) 1 . . . . . . = . . . . . .   ( k ) i =   ( k ) u i z i E ( k ) i S i;c + L ( k ) i;ac (5) By summing the equations abo v e, a closed-form for the update rule of the algorithm can be obtained.   ( k ) i = E 0 X m 2 ~ i z m E ( k ) m S m;c + L ( k ) m;ac (6) F or instance, consider a radial netw ork with 5 nodes and connections f 0–1, 1–2, 2–3, 1–4 g . F or this netw ork, the closed-form for the update rules are the follo wing:   ( k ) 1 = E 0 z 1 E ( k ) 1 S 1 ;c + L ( k ) 1 ;ac (7)   ( k ) 2 = E 0 z 1 E ( k ) 1 S 1 ;c + L ( k ) 1 ;ac z 2 E ( k ) 2 S 2 ;c + L ( k ) 2 ;ac (8)   ( k ) 3 = E 0 z 1 E ( k ) 1 S 1 ;c + L ( k ) 1 ;ac z 2 E ( k ) 2 S 2 ;c + L ( k ) 2 ;ac z 3 E ( k ) 3 S 3 ;c + L ( k ) 3 ;ac (9)   ( k ) 4 = E 0 z 1 E ( k ) 1 S 1 ;c + L ( k ) 1 ;ac z 4 E ( k ) 4 S 4 ;c + L ( k ) 4 ;ac (10) and, in a compact matrix notation: 2 6 6 6 4   ( k ) 1   ( k ) 2   ( k ) 3   ( k ) 4 3 7 7 7 5 = 2 6 6 4 E 0 E 0 E 0 E 0 3 7 7 5 2 6 6 6 6 4 z 1 E ( k ) 1 0 0 0 z 1 E ( k ) 1 z 2 E ( k ) 2 0 0 z 1 E ( k ) 1 z 2 E ( k ) 2 z 3 E ( k ) 3 0 z 1 E ( k ) 1 0 0 z 4 E ( k ) 4 3 7 7 7 7 5 2 6 6 6 4 S 1 ;c + L ( k ) 1 ;ac S 2 ;c + L ( k ) 2 ;ac S 3 ;c + L ( k ) 3 ;ac S 4 ;c + L ( k ) 4 ;ac 3 7 7 7 5 (11) This e xample suggests the update rule can be written in a general compact matrix notation. F or t his accomplishment, assuming an inde xing where u i < i , let us define the lo wer triangular path matrix T with size n n and entries t im = 1 ; if m 2 ~ i 0 ; otherwise (12) Then, the update rule of the algorithm can written in the compact form ( k ) = E 0 TZ p S c K ( k ) TZ p L ( k ) ac K ( k ) (13) where ( k ) is a n 1 v ector with entries gi v en by   ( k ) i , E 0 is a n 1 v ector with entries equal to E 0 , Z p is the n n primiti v e im pedance matrix, S c denotes a n n diagonal matr ix with elements equal to S i;c , L ( k ) ac represents a n n diagonal matrix with elements gi v en by L ( k ) i;ac and K ( k ) is a n 1 v ector with reciprocals of E ( k ) i . Let no w R be a closed re gion in the comple x space C n defined by R = f E 2 C n ; jj E i jj ( E 0 ) , 8 i = 1 ; : : : ; n g , for some 2 R such that E 0 2 < E 0 p (14) where = jj TZ p S c jj + jj TZ p jj L ac ; (15) IJECE V ol. 6, No. 6, December 2016: 3276 3282 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 3279 and L ac ; is a n n diagonal with entries gi v en by the maximum accumulated losses do wnstream each node, computed using a backw ard process with v oltage magnitudes equal to ( E 0 ) . Gi v en an iterate E ( k ) 2 R , by (13) we ha v e ( k ) E 0 = TZ p S c K ( k ) + TZ p L ( k ) ac K ( k ) jj TZ p S c jj K ( k ) + jj TZ p jj L ( k ) ac K ( k ) jj TZ p S c jj ( E 0 ) + jj TZ p jj L ac ; ( E 0 ) (16) and, by using (14) we ha v e ( k ) E 0 ( E 0 ) < ( E 0 ) (17) Therefore, assuming the e xistence of R , for a gi v en if E ( k ) 2 R , then ( k ) belongs to an open ball (in C n ) centered in E 0 and ra d i us equal t o . P articul arly , ( k ) 2 R for all E (0) 2 R , 8 k , then if the initial estimate belongs R , all other iterates also belong to R . 3.2. Con v er gence Rate Once a re gion where iterates are confided through the iterati v e process ha v e be en deduced, let us e xamine the con v er gence rate of the algorithm. Notice that tw o subsequent iterates of the algorithm can be written as   ( k +1) i = E 0 X m 2 ~ i z m E ( k +1) m S m;c + L ( k +1) m;ac (18)   ( k ) i = E 0 X m 2 ~ i z m E ( k ) m S m;c + L ( k ) m;ac (19) and, subtracting (19) from (18) we ha v e   ( k +1) i = X m 2 ~ i z m E ( k ) m S m;c + L ( k ) m;ac z m E ( k +1) m S m;c + L ( k +1) m;ac (20) Hence, by manipulating the terms of the equation abo v e, we ha v e   ( k +1) i = X m 2 ~ i z m " S m;c + L ( k +1) m;ac E ( k +1) m E ( k ) m L ( k +1) m;ac E ( k +1) m E ( k +1) m # E ( k +1) m (21) where E ( k +1) m = E ( k +1) m E ( k ) m and L ( k +1) m;ac = L ( k +1) m;ac L ( k ) m;ac . In a compact form, this e xpression can be re written as   ( k +1) i = X m 2 ~ i D ( k +1) im E ( k +1) m (22) in which D ( k +1) im = X m 2 ~ i z m " S m;c + L ( k +1) m;ac E ( k +1) m E ( k ) m L ( k +1) m;ac E ( k +1) m E ( k +1) m # (23) Equation (22) can also be written in matrix notation as  ( k +1) = D ( k +1) E ( k +1) (24) where  ( k +1) = ( k +1) ( k ) , D ( k +1) is a n n matrix with entries gi v en by D ( k +1) im and E ( k +1) = E ( k +1) E ( k ) . Matrix D ( k +1) indicates the con v er gence rate of the algorithm in each iteration. It also pro v es con v er gence direct dependence on netw ork loading, losses iterates and con v er gence of losses. Con ver g ence Evaluation of a Load Flow Method based on Cespedes’ Appr oac h ... (Die go Issicaba) Evaluation Warning : The document was created with Spire.PDF for Python.
3280 ISSN: 2088-8708 Figure 2. Schematic of the actual distrib ution netw ork utilized in the e v aluations. 4. NUMERICAL RESUL TS Numerical load flo w analysis are presented in this section to v erify the pro vided mathematical results. Fig. 2 sho ws an actual 13.80 kV distrib ution netw ork with 490 nodes utilized to this purpose. W ithout loss of generality , the v alue of 8 : 28 kV (0.6000 pu) has been chosen as . This implies in a feasible re gion R , E 2 C 489 ; jj E i jj E 0 ; 8 i = 1 ; : : : ; 489 g , which meets the interv al 0 : 5000 pu = E 0 2 < E 0 p = 0 : 8160 pu (25) where equals 0.0339 pu. Comple x v oltages ha v e been obtained using the load flo w method. Results ha v e been v alidated using the approach proposed in [1]. T able 1(a) sho ws the real and imaginary parts of v oltage iterates as well as error v alues for node 300, assuming the uncommon initi al estimate of 2 : 00 \ 63 : 03 o pu, 8 i . On the other hand, T able 1(b) sho ws the same v ariables, b ut for a case considering a loading increased by se v enfold. The maximum absolute mismatch between v oltage iterates has been chosen as con v er gence criterion. The last iterate has been set as solution for the sak e of error computation. T olerance has been specified to 10 6 . T able 1. Numerical results for b us 300 assuming the initial estimate E (0) i = 2 : 00 \ 63 : 03 o pu, 8 i (a) Normal loading k e 300 (pu) f 300 (pu) D ( k ) Error 0 0.90719 -1.78241 - - 1 0.99995 -0.01043 0.00013 1.79E 00 2 0.98115 -0.00927 0.00073 2.28E 02 3 0.98096 -0.00936 0.00069 3.16E 04 4 0.98096 -0.00936 0.00079 2.91E 06 5 0.98096 -0.00936 0.00074 2.18E 08 6 0.98096 -0.00936 0 0 (b) Increased loading k e 300 (pu) f 300 (pu) D ( k ) Error 0 0.90719 -1.78241 - - 1 0.99915 -0.07408 0.00677 1.86E 00 2 0.85611 -0.06065 0.05374 1.94E 01 3 0.84061 -0.06560 0.07107 2.48E 02 4 0.83888 -0.06560 0.07926 2.89E 03 5 0.83870 -0.06563 0.07756 3.13E 04 6 0.83868 -0.06563 0.07852 3.20E 05 7 0.83868 -0.06563 0.07816 3.23E 06 8 0.83868 -0.06563 0.07826 2.96E 07 9 0.83868 -0.06563 0 0 Fig. 3 sho ws the first tw o v oltage iterates for the first case in a le v el curv e of the con v er gence re gion. As e xpected, one can notice that e v en by choosing a nonrealistic solution as initial estimate, the first iterate belongs to an open ball centered in E 0 and limited by radius , follo wed t hat the con v er gence of the algorithm is reached. Con v er gence rate is s ho wn to be, in these cases, lo wer to the unit. Furthermore, as deduced, the increase of loading caused an increa se of con v er gence rate, impacting on the algorithm ef ficienc y . IJECE V ol. 6, No. 6, December 2016: 3276 3282 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 3281 e 3 0 0 [ p u ] -2 -1 0 1 2 f 3 0 0 [ p u ] -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 0 1 Solution Voltage iterate E 0 α Figure 3. Iterates in a le v el curv e of the re gion R , for the b us 300. Initial estimate: E (0) i = 2 : 00 \ 63 : 03 o pu. 5. CONCLUSIONS This paper e v aluates the con v er gence of a load flo w method inspired on R. G. Cespedes’ r ecursi v e equations. Numerical results and formal deduction of con v er gence rate sho w that the ef ficienc y of the method depends on the netw ork loading, losses and con v er gence of losses. Also, a re gion where algorithm iterates are confined is deduced as long as initial estimates are chosen properly . Future w orks will e xtend these de v elop- ments to three-phase load flo w approaches. A CKNO WLEDGMENTS The authors w ould lik e to ackno wledge the financial, technical and human support of the CNPq, CAPES, INESC Porto and INESC P&D Brasil. REFERENCES [1] D. Issicaba and J. Coelho, “Rotational load flo w method for radial distrib ution systems, International J ournal of Electrical and Computer Engineering (IJEPE) , v ol. 6, no. 3, 2016. [2] H. Maref atjou and M. Sarvi, “Distrib uted generation allocation to impro v e steady state v oltage stability of distrib ution netw orks using imperialist competiti v e algorithm, International J ournal of Applied P ower Engineering (IJ APE) , v ol. 2, no. 1, pp. 15–26, 2013. [3] J. F . Chen and W . M. W ang, “Uniqueness of the feasible v oltage solutions for radial po wer netw orks, in IEEE Re gion 10 International Confer ence on Micr oeletr onics and VLSI, TENCON’ s 95 , No v emeer 1995, pp. 351–354. [4] H. D. Chiang and M. E. Baran, “On the e xistence and uniqueness of load flo w solution for radial distrib ution netw orks, IEEE T r ansactions on Cir cuits and Systems , v ol. 37, no. 3, pp. 410–416, March 1990. [5] K. N. Miu and H. D. Chiang, “Existence, uniqueness, and monotonic properties of the feasible po wer flo w solution for radial three-phase distrib ution netw orks, IEEE T r ansactions on Cir cuits and Systems I:Fundamental Theory and Applications , v ol. 47, no. 10, pp. 1502–1514, October 2000. [6] E. Bompard, E. Carpaneto, G. Chicco, and R. Napoli, “Con v er gence of the backw ard-forw ard sweep method for the load flo w analysis of radial distrib ution systems, Electrical P ower & Ener gy Systems , v ol. 22, no. 7, pp. 521–530, October 2000. [7] D. Issicaba, “Ladder load flo w methods to radial and weakly meshed dis trib ution systems, Master’ s thesis, Federal Uni v ersity of Santa Catarina, 2008. [8] R. G. Cespedes, “Ne w method for the analysis of distrib ution netw orks, IEEE T r ansaction on P ower Delivery , v ol. 5, no. 1, pp. 391–396, January 1990. Con ver g ence Evaluation of a Load Flow Method based on Cespedes’ Appr oac h ... (Die go Issicaba) Evaluation Warning : The document was created with Spire.PDF for Python.
3282 ISSN: 2088-8708 BIOGRAPHIES OF A UTHORS Diego Issicaba recei v ed the B.S. and M.S. de grees in Electrical Engineering from the Federal Uni- v ersity of Santa Catarina (UFSC), Santa Catarina, Brazil, in 2006 and 2008, respecti v ely . Fur - thermore, he recei v ed the Ph.D. de gree on Sustainable Ener gy Systems, under the MIT Doctoral Program, from the F aculty of Engineering of the Uni v ersity of Porto, Portug al. His research inter - ests in v olv e smart grids, mutiagent systems, distrib uted generation and distrib ution systems. He is currently a full Professor at Federal Uni v ersity of T echnology P arana (UTFPR), Associate Re- searcher and Coordinator of the Research Area on Ener gy and Management of INESC P&D Brasil. J or ge Coelho recei v ed the B .S. and M.S. de grees in electrical engineering from the Federal Uni v er - sity of Santa Catarina, Brazil, in 1977 and 1980, respecti v ely . In 1990, he recei v ed the Ph.D. de gree in electrical engineering from the Catholic Uni v ersity of Rio de Janeiro, Brazil. He is a Professor of the Department of Electrical Engineering at the Federal Uni v ersity of Santa Catarina, Brazil, since March 1978. His research interests include distrib ution systems e xpansion and opera tion planning, po wer systems reliability , probabilistic methods applied to po wer systems, and po wer quality . IJECE V ol. 6, No. 6, December 2016: 3276 3282 Evaluation Warning : The document was created with Spire.PDF for Python.