3) © (8) Fig. 3.0 Power sys- tem representations. ja) Single line dia- gram; (b) positive se- quence network dia- gram;(c)oriented con- @ nected graph.Qe . » Branch ~~ Link a ins Fig. 3.2 Tree and cotree of the oriented connected graph A connected subgraph containing all nodes of ¢ graph hut no closed path is called a tree, The elements of a tree are called branches and furni a subset of the elements of the connected graph. ‘The number of branches b required to form a tree is b n-1 (3.2.1)A connected subgraph containing ell nodes of a graph hut no elosed path is called a free, The elements f a tree are called branches and furn, asubset of the elements of the connected graph. The number of branches b required to form a tree is b=n-1 (3.2.1) where n is the number of nodes in the graph. Those elements of the connected graph that are not included in the tree are called links and form a subgraph, not necessarily connected, called the cotree. The cotree is the complement of the tree. The number of links l of a connected graph with ¢ elements is l=e-bd From equation (3.2.1) it follows that lee-ndl B22 A tree and the corresponding eotree of the graph given in Fig. 3.l¢ are shown in Fig. 3.2 Ifa link is added to the tree, the resulting graph contains one elo path, called a loop. The addition of each subsequent link forms one «r more additional loops. Loops which contain only one link are inde- pendent and are called basic loops. Consequently, the number of basic 4@ Fig. 3.3 Basic loops of the oriented connected graph. loops is equal to the number of links given by equation (8.2.2). Orienta~ tion of a basic loop is chosen to be the same as that of its link. ‘The basic loops of the graph given in Fig, 3.2 are shown in Fig.3.3 Incidence matrices Element-node incidence matrix A The incidence of elements to nodes in a connected graph is shown by the element-node incidence matrix. The elements of the matrix are as follows: a, = 1 if the ‘th element is incident to node 1 if the ith element is incident to and wrtented toward the jth node a, = Oif the rth element is not incident to the th nade nd oriented away from the jth The dimension of the matrix is ¢ X n. where «18 the number of elements and n is the number of nodes in the graph. ‘The element-node incidence matrix for the graph shown in Fig. 3.2 isThis matrix is reetangular and therefore singular. If the rows of A are arranged according to a paz cular tree, the matrix can be partitioned into submatrices Ay of dimension b X (n — 1) and Ar of dimension £ X (mn — 1), where the rows of Ay correspond to branches and the rows of .1; to links. The partitioned matrix for the graph shown in lig. 3.2 isN Buses Ay Links Ap is a nonsingular square matrix with rank (n ~ i}7.5 THE NETWORK INCIDENCE MATRIX AND Yous In Secs, 7.1 and 7.2 nodal admittance equations for each branch and mutually coupled pair of branches are derived independently from those of other branches in the network. The nodal admittance matrices of the individual branches are then combined together in order to build Y,y, of the overall system. Since we now understand the process, we may proceed to the more formal approach which treats all the equations of the system simultaneously rather than sepa- rately. We will use the example system of Fig. 7.11 to establish the general procedure. Two of the seven branches in Fig. 7.11 are mutually coupled as shown. The mutually coupled pair is characterized by Eq. (7.14) and the other five=j080 - : . : : : ve] [le -j6.25 j3I5 : : : v, I, BIS -j625 - - : . v, I. —j8.00 Velo | le ~j5.00 be I, -j2.50 vy, ly —j0.80 |] Y% I, (7.28) The coefficient matrix is the primitive admittance matrix formed by inspection of Fig. 7.11. Each branch of the network contributes @ diagonal entry equal to the simple reciprocal of its branch impedance except for branches 6 and c, which are mutually coupled and have entries determined by Eq. (7.13). For the general case Eq. (7.28) may be more compactly written in the form Y,Voe = pe pr (7.29)where Vp, and I,, are the respective column vectors of branch voltages and currents, while Y,, represents the primitive admittance matrix of the network. The primitive equations do not tell how the branches are configured within the network. The geometrical configuration of the branches, called the topology, is provided by a directed graph, as shown in Fig. 7.13(a) in which each branch of the network of Fig. 7.11 is represented between its end nodes by a directed line tree branch linksegment with an arrow in the direction of the branch current. When a branch connects to a node, the branch and node are said to be incident. A tree of a graph is formed by those branches of the graph which interconnect or span all - the nodes of the graph without forming any closed path. In general, there are many possible trees of a network since different combinations of branches can be chosen to span the nodes. Thus, for example, branches @, b, c, and f in Fig. 7.13(b) define a tree. The remaining branches d, e, and g are called links, and when a link is added to a tree, a closed path or loop is formed. A graph may be described in terms of a connection or incidence matrix. Of particular interest is the branch-to-node incidence matrix A, which has one row for each branch and one column for each node with an entry @,; in row i and column j according to the following rule: 0 if branch ¢ is not connected to node G) ={ 1 ifcurrent in branch / is directed away from node @)_ (7-30) =1 if current in branch / is directed toward node (J)This rule formalizes for the network as a whole the procedure used to set up the coefficient matrices of Eqs. (7.6) and (7.15) for the individual branches. In network calculations we usually choose a reference node. The column corre- sponding to the reference node is then omitted from A and the resultant matrix is denoted by A. For example, choosing node (@) as the reference in Fig. 7.13 and invoking the rule of Eq. (7.30), we obtain the rectangular branch-to-node matrix a > d O} 0 0 -1 -1 a -1 0 @@® 0 —1 1 1 1 (7.31) wo escc®he voltage across each branch may be expressed as the difference in its énd-bus voltages measured with respect to the reference node. For example, in Fig. 7.11 the voltages at buses (1), @), @, and @) with respect to reference node (@) are denoted by V,, V;, V3, and V4, respectively, and so the voltage drops across the branches are given by uy . oo 0 1 MY, oO -1 1 Of, Ke -1 9 1 of Ve or Weye}-1 1 0 off? v, Vv. 0 -1 0 ify 4 ¥, -1 0 o aft *J Weak ¥, o 0 0 1] in which the coefficient matrix is the A matrix of Eq. (7.31). This is one illustration of the general result for any N-bus network given by Vv. pe =av . (7.32)where V,, is the B X1 column vector of branch voltage drops and V is the Nx 1 column vector of bus voltages measured with respect to the chosen reference node. Equations (7.6) and (7.15) are particular applications of Eq. (732) to individual branches. We further note that Kirchhoff’s current law at nodes (1) to () of Fig. 7.11 yields Ja a) 4s 0 0 -1 -1 0 =1 Off 0 Oo -1 0 41 -1 0 Oo] f]_]o a\= 1 1 1 0 0 0 Oo} 1 0 0 0 o 1 4 aie Is f , where J, = 1.00/ —90° and J, = 0.68, 135° are the external currents in- jected at nodes (3) and (), respectively. The coefficient matrix in this equation “is AT. Again, this is illustrative of a general result applicable to every electrical network since it simply states, in accordance with Kirchhoff’s current law, that the sum of all the branch currents incident to a node of the network equals theinjected current at the node. Accordingly, we may write 7 ATL, = 1 (7.33) where I,, is the B 1 column vector of branch currents and I is the NX 1 column vector with a nonzero entry for each bus with an external current source. Equations (7.5) and (7.16) are particular examples of Eq. (7.33) ‘The A matrix fully deseribes the topology of the network and is indepen- dent of the particular values of the branch parameters. The latter are supplied by the primitive admittance matrix. Therefore, two different network configura tions employing the same branches will have different A matrices but the same Y,,. On the other hand, if changes occur in branch parameters while maintain- ing the same network configuration, only Y,, is altered but not A. Multiplying Eq. (7.29) by A, we obtain AYYV a, = ATL (734) The right-hand side of Eq. (7.34) equals I and substituting for V,, from Eq. (7.32), we find {AvY,,A}]V =We may write Eq. (7.35) in the more concise form Y,,.V = 1 (7.36) where the W x N bus admittance matrix ¥,,, for the system is given by Yo = a Yn OA (737) a NXN NXB BxXB BxXN Yj. has one row and one column for each of the V buses in the network, and so the standard form of the four independent equations of the example of Fig. 7.11 is system Yr Ya Yo YalfM] [4 Yo Yoo Yo Yau || ¥2 h Boe a Le (7.38) Yy Yoo Yas Yuu | I Ya Ya Ya Yea |[M fp [a]Example 7.5. Determine the per-unit bus admittance matrix of the example system of Fig. 7.11 using the tree shown in Fig. 7.13 with reference node (0). Solution. The primitive admittance matrix Y,, describing the branch admittances is given by Eq. (7.28) and the branch-to-node incidence matrix A for the specified tree is given by Eq (7.31). Therefore, performing the row-by-column multiplications for A”Y,, indicated by 7 0 0 1 of f-j08 o -1 10 =16.25 13.75 -1 010 BIS 6.25 -1 100 : -j8.0 Oo -101 5.0 -1 o1 j2.5 Lo 001 -j08 we obtain the intermediate result 0 =/3.75 j6.25 j8.0 0 725 0 ary 0 j625° -/3.75 -j8.0— js.0 0 0 or -j08 -j25 -j25 0 0 0 0 0 0 0 0 ~j50 -j25 -j08which we may now postmultiply by A to calculate ® @ @® | -/16.75 SULTS J2.50 j2.50 @I| ju7s -j19.25 * j2s0 5.00 @] i250 32.50 ~J5.80 0 @| i250 5.00 0 78.30 Yous = AT Y yA = Since currents are injected at only buses @) and @), the nodal equations in matrix form are written -j16.75 fll.7S 250 j2.80 | Y, 0 ju7s =j19.25 3250 js.00 || v, 0 j250 j250 —j5.80 0 |) v3) =] 1.007 —90° j2.50 j5.00 0 ~j8.30 || % 0.68/ —135°7.6 THE METHOD OF SUCCESSIVE ELIMINATION In industry-based studies of power systems the networks being solved are geographically extensive and often encompass many hundreds of substations, generating plants and load centers. The Y,,, matrices for these large networks of thousands of nodes have associated systems of nodal equations to be solved for a correspondingly large number of unknown bus voltages. For such solutions computer-based numerical techniques are required to avoid direct matrix inver- sion, thereby minimizing computational effort and computer storage. The method of successive elimination, called gaussian elimination, underlies many of the numerical methods solving the equations of such large-scale power systems. We now describe this method using the nodal equations of the four-bus system Yuk + Yiqbs + Yiaks + Yuh = hh (7.39) YoY, + Yaa + Yaa + YoMe (7.40) Yah, + Yoho + Yaa + Yue = ly (7.41) Yul, + Yas + Yas + Yaals = la (7.42)Gaussian elimination consists of reducing this system of feur equations in the four unknowns Vy, Vs, Vy, and Vy to a system of three equations in three unknowns and then to a reduced system of two equations in two unknowns, and so on until there remains only one equation in one unknown. The final equation yields a value for the corresponding unknown, which is then substituted back in the reduced sets of equations to calculate the remaining unknowns. The successive elimination of unknowns in the forward direction is called forward elimination and the substitution process using the latest calculated values is 264°) CHAPTER? THE ADMITTANCE MODEL AND NETWORK CALCULATIONS called back substitution. Forward elimination begins by selecting one equation and eliminating from this equation one variable whose coefficient is called the pivot. We exemplify the overall procedure by first eliminating V, from Eqs. (7.39) through (7.42) as follows:Step 1 1. Divide Eq. (7.39) by the pivot ¥,, to obtain y, y, y+ Py + (7.43 vit yy ) 2, Multiply Eq. (7.43) by Ya), Ys, and Y,, and subiract the results from Eqs. (7.40) through (7.42), respectively, to get vn Vy 4 [ry — 2% Yv, [ry — BM PEE a Py,Equations (7.43) through (7.46) may be written more compactly in the form Ya Yo, Ya Wet Sht swt Ya Ya fy yy! YPr + YI + YN, = LP YPWa + YIQVs + YW, = 1S? i YP, + YEP + YEPY, = 1 ‘where the superscript denotes the Step 1 set of derived coefficients yo YMue =2,3,4 YP =¥e- EE for jand k= 2,3, n (7.47) (7.48) (7.49) (7.50) (7.51)and the modified right-hand side expressions for j= 2,3,4 (7.52) Note that Eqs. (7.48) through (7.50) may now be solved for V,, V,, and V, since V, has been eliminated; and the coefficients constitute a reduced 3 x 3 matrix, which can be interpreted as representing a reduced equivalent network with bus @ absent. In this three-bus equivalent the voltages /,, y, and V, have exactly the same values as in the original four-bus system. Moreover, the effect of the current injection /, on the network is taken into account at buses @), @), and @), as shown by Eq. (7.52). The current /, at bus (1) is multiplied by the factor =¥n/¥; before being distributed, so to speak, to each bus (J) still in the network. We next consider the climination of the variable V:Step 2 1. Divide Eq. (7.48) by the new pivot Ys to obtain yn Y 2 ey. yo a3 7 Yn ys? ¥, gp (7.53) ‘4 3 - 2. Multiply Eq. (7.53) by YP and ¥{? and subtract the results from Eqs. (7.49) and (7.50) to get vQvse [> - ype [P@ fooy YOR [yee “le yA — yyry ¥p : = 10 — 221 (7.54) YP oo yp nn) wy 1 — S20) (7.55) 7 pe? In a manner similar to that of Step 1, we rewrite Eqs. (7.53) through (7.55) in the formIn a manner similar to that of Step 1, we rewrite Eqs. (7.53) through (7.55) in the form yy yp I ie My Ve y's + yh yor (7.56) YPV, + YSPV, = 1? (7.57) YPV, + YQ, = 1 * (7.58)where the second set of calculated coefficients is given by for jand k = 3,4 (7.59) 2 pn — 728 yay 1 = If If (7.60) Equations (7.57) and (7.58) describe a further reduced equivalent network having only buses @) and @). Voltages V, and V4 are exactly the same as those of the original four-bus network because the current injections 12) and (2 represent the effects of all the original current sources. We now consider elimination of the variable V5. Step 3 1. Divide Eq, (7.57) by the pivot ¥{? to obtain (7.61)Step 3 1. Divide Eq. (7.57) by the pivot ¥{? to obtain Yo Vs ys (7.61) 2. Multiply Eq. (7.61) by Y,9) and subtract the result from Eq. (7.58) to obtain YQ, = 1 (7.62) in which we have defined YPYP YY? ; 2 5 ¥{P = YIP - P and = [9 = 12) yal? (7.63) Equation (7.62) describes the single equivalent branch admittance ¥, with voltage V, from bus @) to reference caused by the equivalent injected current 12. The final step in the forward elimination process yields Vy.Step 4 1. Divide Eq. (7.62) by ¥,2 to obtain 1 Y= ya (7.64) ‘se At this point we have found a luc for bus voltage 4, which can be substituted 2.6 THE METHOD OF SUCCESSIVE ELIMINATION 267 back in Eq. (7.61) to obtain a value for V,. Continuing this process of back substitution using the values of ¥, and V, in Eq, (7.56), we obtain V, and then solve for Yo fram Fa (747)Thus, the gaussian-elimination procedure demonstrated here for a four-bus system provides a systematic means of solving large systems of equations without having to invert the coefficient matrix. This is most desirable when a large-scale power system is being analyzed. The following example numerically illustrates the procedure. Example 7.7. Using gaussian elimination, solve the nodal equations of Example 7.5 to find the bus voltages. At each step of the solution find the equivalent circuit of the reduced coeflicient matrix. Solution. In Example 7.5 the nodal admittance equations in matrix form are found tobe oO ® ® . @ [ ~i16.75 dL.7S j2.50) |" 0 @| (FB) -j19.25 i500] M2 | ° @| i250 52.50 0.00] |% 1,00/ — 90° @| 2-50 75.00 0.00 ~18.30 | [Ms 0.687 — 135°Step 1 To climinate the variable V, from rows 2, 3, and 4, we first divide the first row by the pivot —j16.75 to obtain V, — 0.70149V, — 0.14925V, ~ 0.14925V, = 0 We now use this equation to eliminate the j11.75 entry in (row 2, column 1) of Yjos and in the process all the other elements of row 2 are modified. Equation (751) shows the procedure. For example, to modify the element j2.50 under- lined in (row 2, column 3), subtract from it the product of the elements enclosed by rectangles divided by the pivot —j16.75; that is, Yah iy oy — TTS L250 a303 . a ee t yo ae! per unil VV = Yo Similarly, the other clements of new row 2 JALTS X j1.75 = 19.25 — = ~/11.00746 per unit YgP = =j19.25 sas j Pe F175 x j2.50 . = Sage 77679373 per unit~j4.25873 ~§6.75373 1.00/- 90° (4 98CAY 668/138 FIGURE 7.14 ‘The equivalent three-bus network following Step 1 of Example 7.7 Modified elements of rows 3 and 4 are likewise found to yield= 0.70149 -0.14925 ~0.14925 ][-V, 0 =11.00746—4.25373——j6.75373 || V, 0 1 0 = | j4.25373 -j5.42686 © 0.37313 |] Vy 1.00/ -90° 0 J6.75373 j0.37313 —j7.92686 || V, 0.68/ ~ 135° | Because J; = 0, no current is distributed from bus (1) to the remaining buses , 3), and @), and so the currents /!, /1, and 7{ in the right-hand-side vector have the same values before and after Step 1. The partitioned system of equations involving the unknown voltages V>, V,, and V, corresponds to the three-bus equivalent network constructed in Fig. 7.14 from the reduced coeffi- cient matrix. Step 2 Forward elimination applied to the partitioned 3 x 3 system of the last equa- tion proceeds as in Step 1 to yield 1 0.38644 — 0.61356 || ¥%2 0 1 —0.70149 | — 0.14925 — 0.14925 ][ 0 0 0 0 | ose: 72.98305 |} V5 1.00/ —90° . Q | 0 2.98305 —j3.78305 || V, 0.68, 135°® ~j08 "(h) 0068/2138 FIGURE 7.15 ‘The equivalent two-bus network following Step 2 of Example 7.7. in which the injected currents /{ and 1% of Step 2 are also unchanged because I$” =/, = 0. At this stage we have eliminated V, and V, from the original 4 x 4 system of equations and there remains the 2 x 2 system in the variables V, and V, corresponding to Yyy, of Fig. 7.15. Note that nodes (1) and @) are eliminated.Step 3 Continuing the forward elimination, we find 1 -0.70149 —0.14925 | — 0.14925 I[ V, ] 0 0 1 —0.38644 | —0.61356 || V, 0 0 0 1 —0.78853 |] V, | = 0.26434 / 0° 0 0 Oo | 7143082 1.357387 — 110.7466" in which the entry for bus (3) in the right-hand-side vector is calculated to be 12 1,00, 90° = 0.264340" per unit Ys? 3.783058/ — 90° and the modified current at bus (@) is given by @ 1217-2 a PIP yoyls § 0.68 / — 135° 298505 1.00 / — 90° oe ~ 3.78305 —<——=1 35738/ — 110.7466 per unit Figure 7.16(a) shows the single admittance resulting from Step 3. 270 CHAPTER? THE ADMITTANCE MODEL AND NETWORK CALCULATIONS + rer —| ® 4 1.43082 | + _) osseor 20.7865, 1.35738/— 110.7466" Jot (a) (6)Step 4 The forward-elimination process terminates with the calculation ye 1 4.35738/ ~ 110.7466 “Ye? 1.83082/ -90° corresponding to the source transformation of Fig. 7.16(b). Therefore, forward elimination leads (o the triangular coefficient matrix given by 466° ~ = 0.94867/ - 20.7466" per unit 1 0.70149 —0.14925 -0.14925 ] [ 1 0 0 1 —0.38644 —0.61356 |} 4 0 0 0 1 —0.78853 || v5 | 0.26434 / 0° 0 0 0 1 | yy | 0.94867/ — 20.7466" Since V, = 0.948677 — —20.7466°, we now begin the back-substitution process to . determine I’, using the third row entries as follows: V, — 0.78853, = Vy — 0.74805 / —20 7466" = 0.26434 / 0° which yields¥, = 0.99965 / - 15.3716" per unit Back substitution for V, and V, in the second-row equation V, — 0.386447; — 0.61356V, = 0 yields V, = 0.96734/ ~ 18.6030 per unit . 7.7 NODE ELIMINATION (KRON REDUCTION) 271 When the calculated values of V2, 13, and V, are substituted in the first-row equation . V, — 0.70149V, — 0.14925V, — 0.14925V, = 0we obtain V, = 0.96903 / ~ 18.4189" per unit and so the per-unit bus voltages are Vv, = 0.96003/ - 18.4189° = 0.91939 — j0.30618 V7, = 0.96734 / - 18.6030" = 0.91680 — j0.30859 V, = 0.99904 / —15.3716° = 0.96388 - J0.26499 V, = 0.94867/ —20.7466° = 0.88715 — j0.33605 which agree almost exactly with the results found in Example 7.6. 7.7 NODE ELIMINATION (KRON REDUCTION)