IEEE Transactions on Power Delivery, Vol. 8, No. 4, October 1993 1890 A COMPARISON AMONG EIGHT DIFFERENT TECHNIQUES TO ACHIEVE AN OPTIMUM ESTIMATION OF ELECTRICAL GROUNDING PARAMETERS I N TWO-LAYERED EARTH J.L. d e l Alamo, E . E . , Ph.D., I E E E Member Dept. o f E l e c t r i c a l Engineering U n i v e r s i t y of V a l l a d o l i d SDain ABSTRACT.-This paper intends t o compare t h e performance o f e i g h t techniques, based on f i v e d i f f e r e n t methods, used t o f i n d t h e e l e c t r i c a l grounding parameters o f a two-layered earth (resistivities and thickness) corresponding to a specific mathematical model. Parameter e s t i m a t i o n i s c a r r i e d o u t - i n such a way as t o g e t an optimum f i t t i n g between t h e s e t of r e s i s t i v i t y values measured i n f i e l d by means o f Wenner’s Method, and those c a l c u l a t e d from t h e mathematical model u s i n g such parameters. F i r s t l y , t h e paper i s devoted t o p r e s e n t i n g a review o f t h e mathematical b a s i s o f f i v e o p t i m i z a t i o n methods used f o r t h e implementation o f t h e d i f f e r e n t techniques s u b j e c t t o comparison. Secondly, t h e paper i s dedicated t o d e f i n i n g t h e implementation o f t h e a l g o r i t h m s f o r the eight techniques. Three o f them a r e o f F i r s t Order Gradlent type, one more o f Second Order (Newton Method), t h e f i f t h one i s based on t h e Levenberg-Marquardt Method, t h e s i x t h one on t h e Generalized I n v e r s e Method, t h e seventh one on a Quasi-Newton Method and t h e l a s t one on a mixed Newton-Generalized I n v e r s e method. Algorithms a r e a p p l i e d t o s i x t e s t cases comparison o f t h e r e s u l t s i s shown. Furthermore, algorithms to improve existing procedures presented. and new are V pa(xP,r , ) Gradient of t h e f u n c t i o n mathematically J m o d e l l i n g t h e ground w i t h t h e parameters 5 , evaluated a t p o i n t P o r i n s t e p p. x - xp A 6 F(x) n i vector 3 xp from = O b j e c t i v e Function f o r l e a s t squares f i t t i n g between measured values and computed values, according t o a s p e c i f i c mathematical model. O b j e c t i v e Function evaluated a t p o i n t P o r i n F(xp) step P Superindex t o i n d i c a t e t r a n s p o s i t i o n o f a matrix o r vector Residual Vector w i t h m elements o f t h e weighted d i f f e r e n c e s . Each element i s i n t h e (r.) form d j = (pm(rj)-pa(x,r.))/P J m~ = Jacobian M a t r i x w i t h p a r t i a l d e r i v a t i v e s o f weighted d i f f e r e n c e s w i t h respect t o each parameter. dimension v x m. 1 J . . = -- apa(X, r .)/ax .= ad ./ax. J1 Pm J 1 J 1 t D J V F(xp)= Gradient o f t h e O b j e c t i v e Function evaluated a t point P o r i n step p Dimension v x 1. V F(xp) 2 Jt D n U n i t M a t r i x , dimension H Hessian M a t r i x o f F ( x ) evaluated a t p o i n t o r i n step p, as b r i e f n o t a t i o n o f : LIST OF SYMBOLS [ Upper l a y e r r e s i s t i v i t y , i n 0.m VF(_Xp) 1. v x v Dimension P v x v Lower l a y e r r e s i s t i v i t y , i n Q.m R e f l e c t i o n c o e f f i c i e n t ( p 2 - p , )/(p2+P, 1 Upper l a y e r t h i c k n e s s Vector with parameters d e f i n i n g t h e s o i l . I n general w i t h v elements Xi 3 For two-layer v Probe spacing (F.Wenner), r/z (i-1 J E i n meters. V P2’Pl j J second d e r i v a t i v e s . = Vector o f weighted d i f f e r e n c e s , w i t h j rows. I t s elements (Generalized I n v e r s e Method) are: .....v). Number o f f i e l d measurements. Measured r e s i s t i v i t y (Q.m) f o r each spacing rJ. p,(&,r.)= = M a t r i x o f weighted r e s i d u a l s d . and t h e i r R ..... = 1 m. B r i e f l y pm. Apparent r e s i s t i v i t y (0.m). 11 E ]I2= Gradient o f t h e squared kuclidean Norm o f E i n the Vector o f weighted d i f f e r e n c e s Generalized I n v e r s e Method. INTRODUCTION Computed value u s i n g t h e mathematical model f o r each spacing r . and v e c t o r 8 B r i e f l y pa. J . 93 WM 079-4 PWRD A paper recommended and approved by the IEEE Transmission and Distribution Committee o f the IEEE Power Engineering Society for presentation at the IEEE/PES 1993 Winter Meeting, Columbus, OH, January 31 - February 5, 1993. Manuscript submitted January 31, 1992; made available for printing November 3 , 1992. A t t h e i n i t i a l s t e p o f e l e c t r i c a l grounding design o f a S t a t i o n , Substation, Transformer D i s t r i b u t i o n o r a L i n e support, i t i s necessary t o o b t a i n t h e parameters d e f i n i n g t h e s o i l where such grounding i s going t o be done. The parameter e s t i m a t i o n must always be c a r r i e d o u t based on a s e t o f measurements o f s o i l r e s i s t i v i t y . The i n t e r p r e t a t i o n o f measurements c a r r i e d o u t i n t h e f i e l d through a technique corresponding t o a s p e c i f i c mathematical model o f t h e s o i l c o n s i s t s i n f i n d i n g a s e t o f e l e c t r i c a l grounding parameters d e f i n i n g i t s e l f i n such a way as t o a t t a i n a good fitness, preferably an optimum f i t n e s s , between measured values and those which would r e s u l t u s i n g such parameters i n t h e mathematical model. 0885-8977/93/$03.000 1993 IEEE 1891 The m a j o r i t y of t h e times measurements w i l l i n d i c a t e t h a t t h e s o i l , w i t h i n a l i m i t e d extension, behaves as i f i t was composed [ S I [ 9 ] [12] [15] o f two h o r i z o n t a l layers. 3rd.-We admit t h a t we have a s e t o f r e s i s t i v i t y values p ( r . ) c o l l e c t e d i n f i e l d f o r d i f f e r e n t spacing m~ rj (j 1 .......m), 4th.-The I n t h i s paper we w i l l consider t h a t s o i l measurements have been made through t h e F. Wenner Method, and so t h e mathematical model corresponds t o t h i s Method. Along t h i s , several i n v e s t i g a t o r s have with solutions to the problem of contributed two-layered parameter s o i l e s t i m a t i o n t o achieve t h e b e s t p o s s i b l e f i t between t h e measured and computed E P R I [61, r e s i s t i v i t y values. The c o n t r i b u t i o n s from C.J.Blatner [7], ANSI/IEEE [91, F.Dawalibi and [ l o ] , C.J.Blatner [121, A.P.Meliopoulos C.J.Blatner [13] and, r e c e n t l y , J.L.de1 Alamo [ l a ] should be highlighted. I n [16] a very compact f o r m u l a t i o n f o r t h e a n a l y s i s o f t h e e v o l u t i o n forms o f c a l c u l a t e d apparent r e s i s t i v i t y i n m u l t i - l a y e r s o i l s i s presented, and an e s t i m a t i o n procedure i s c a r r i e d o u t based i n a p - r comparison technique. However no a n a l y t i c a l s o l u t i o n t o t h e t h e problem o f parameter e s t i m a t i o n i n such kinds o f s o i l s i s included. I n references [ll] and [13] e s t i m a t i o n o f s o i l parameters i s made when measurements have been c a r r i e d o u t f o l l o w i n g t h e d r i v e n rod method. I n [12] a comparison o f r e s u l t s i s shown. As i s i n d i c a t e d i n [ 7 ] [13] and 1181, t h e m a j o r i t y o f these techniques a r e n o t very e f f e c t i v e because, even i f t h e s o i l was a b l e t o be shaped as two-layers, t h e a l g o r i t h m s present slow convergence o r have n o t enough s t a b i l i t y t o converge t o t h e s o l u t i o n . However i t has been shown [ l a ] t h a t a second order g r a d i e n t technique i s capable o f a r r i v i n g a t t h e intended absolute minimum o f t h e O b j e c t i v e Function. The research work c o n s t i t u t i n g t h e s u b j e c t o f t h i s paper i s one o f t h e f i r s t r e s u l t s from t h e Research P r o j e c t No. 132197 sponsored by both OCIDE, O f i c i n a de Coordination e I n v e s t i g a c i d n de D e s a r r o l l o E l e c t r o t k c n i c o (Coordinating Bureau o f Research and E l e c t r i c a l Engineering Development), and t h e Company IBERDROLA, i n Spain, t o d e f i n e t h e b e s t t o o l f o r optimal parameter e s t i m a t i o n i n a m u l t i l a y e r s o i l from r e s i s t i v i t y measurements through t h e Wenner Method. METHODOLOGY We wish t o compare t h e e f f i c i e n c y o f d i f f e r e n t methods o f s o i l parameter e s t i m a t i o n f o r t h e optimal f i t between a s e t o f measured r e s i s t i v i t y values i n f i e l d and those which would r e s u l t from t h e a p p l i c a t i o n of a mathematical model u s i n g these parameters. For t h i s o b j e c t i v e we use a homogeneous treatment o f t h e problem, based on t h e f o l l o w i n g hypothesis: 1st.-We admit t h a t t h e s o i l i s composed by two horizontal homogeneous layers or strata, being s e m i - i n f i n i t e , with d i f f e r e n t r e s i s t i v i t y , p1 and P2 , and z being t h e t h i c k n e s s o f t h e upper l a y e r . L e t K = ( p -p )/(p2+pl) the r e f l e c t i o n coefficient. 2 1 2nd.-For t h e computed apparent r e s i s t i v i t y , u s i n g t h e F.Wenner Method as model, w i t h spacing r between probe electrodes, we w i l l use [81 [ l o ] 1121: where A 1 = + l.....m ( 2 n z / r ) 2 and B = A + pm(r1), pm(r2) 6 vector ,.....Pm ( r m1. defining the soil parameters w i l l be used i n each treatment: ? = [z'] or, a l t e r n a t i v e l y f o r the comparison w i t h t h e proposed technique i n [ 9 ] : [?I _X 5h.-We w i l l n o t e t h e weighted d i f f e r e n c e d . J (j = l...m) between measured and computed values f o r a c e r t a i n spacing r j and w i t h a s e t o f parameters 5 , by t h e expression: J t ' t h e weighted d i f f e r e n c e s w i l l c o n s t i t u t e t h e v e c t o r D, w i t h dimension m x 1. All 6th.-We w i l l consider t h a t t h e s e t o f parameters & which b e s t shapes t h e s o i l i s t h a t which optimizes t h e square o f a f u n c t i o n F(X), d e f i n e d as t h e Euclidean ,I1 D 112, o f v e c t o r D. T h i s corresponds t o t h e Norm m a j o r i t y o f t h e adjustment f u n c t i o n s used i n t h e l i t e r a t u r e [61 [ l o ] as: j = t....m METHODS TO BE USED To compare t h e procedures we w i l l implement e s t i m a t i o n techniques based on t h e f o l l o w i n g methods: Steepest descent method Levenberg-Marquardt method Newton method Generalized I n v e r s e method Quasi-Newton method STEEPEST DESCENT METHOD Based on t h i s method, several techniques o f parameter e s t i m a t i o n have been presented, such as those proposed i n EPRI Research P r o j e c t 149-Report EL 2699, References [61 [ l o ] , and ANSI/IEEE Std. 81-1983 Ref.[9]. Also presented i n t h i s paper i s an a u x i l i a r y technique based on t h i s method t o achieve an absolute minimum f o r t h e o b j e c t i v e f u n c t i o n o f e r r o r s between measured and comouted values. I t i s known t h a t t h e negative g r a d i e n t i s i n t h e d i r e c t i o n o f descent step. So approaching t h e minimum w i l l be c a r r i e d o u t through an i t e r a t i v e procedure such that: XP+' = Xp - - A V F(Zp) Where A i s a p o s i t i v e s c a l e f a c t o r . The main problems o f convergence, when u s i n g a technique o f t h e steepest descent method, a r e caused by t h e s c a l e f a c t o r A. To avoid i n p a r t t h e s c a l i n g problem we can s e l e c t an u n i t v e c t o r d e f i n e d as: VF(Xp) / IIF(&p)I( and then t h e process o f approaching t h e minimum w i l l be expressed i n t h e form [181: Pa = P1 ( 1 + 4 n i.e. same 3 1892 Steepest descent method i s a poor search s t r a t e g y . I t u s u a l l y works q u i t e w e l l [ 5 1 d u r i n g e a r l y stages o f t h e o p t i m i z a t i o n process. However, as a s t a t i o n a r y p o i n t i s approached, t h e method u s u a l l y i s i n e f f i c i e n t when small orthogonal steps are taken. F i r s t d e r i v a t i v e s o f t h e O b j e c t i v e Function are needed f o r a l l o f t h e techniques being analyzed. A l l t h e r e q u i r e d i n f o r m a t i o n i s gathered i n t h e Appendix. must It formulation, be noted that according t h e changes i n v e c t o r to from _Xp to P where U zero, the following + [ H i s called CT P ] AXp = minimum c o n d i t i o n - is VF(X P ) t h e Levenberg-Marquardt parameter. For t h e problem we a r e considering, and according t o t h e f o r m u l a t i o n i n t h e Appendix: Ift h e r e s i d u a l s were l i n e a r , then: that 5’’ can be computed through t h e expression: where Lagrangian reached: (4) I t must be noted t h a t , i n a c e r t a i n way [ 1 4 ] , t h e m a t r i x o f r e s i d u a l s R and i t s second d e r i v a t i v e s have been s u b s t i t u t e d by t h e diagonal m a t r i x CT 1. i s a s c a l e f a c t o r s i m i l a r t o X. One o f t h e d i f f i c u l t i e s i n t h e implementation o f t h i s method [14] c o n s i s t s i n c o n t r o l l i n g t h e value o f NEWTON METHOD Considering t h a t around a p o i n t ( v e c t o r ) Xp, parameter the O b j e c t i v e Function i s shaped by means o f a q u a d r a t i c f u n c t i o n , t h e value o f t h e f u n c t i o n a t any p o i n t i n t h e p r o x i m i t y o f t h a t p o i n t can be expressed by: The sequence o f searching o f t h e minimum has t h e form [ 5 ] : AEp = - H-’ VF(Xp) The elements o f t h e Hessian m a t r i x (and those o f t h e Gradient v e c t o r ) are reported i n t h e Appendix. The descent property is paramount for an o p t i m i z a t i o n a l g o r i t h m , t h a t i s , each s t e p must proceed d o w n h i l l [14]. I n Newton’s method t h i s i s p o s s i b l e , f o r any AX i f and o n l y i f t h e Hessian i s p o s i t i v e definite. Newton’s Method, i n general, does n o t converge t o a s t a t i o n a r y p o i n t [ 5 ] when we s t a r t w i t h an a r b i t r a r y i n i t i a l p o i n t . For t h e problem we are considering, and according t o t h e f o r m u l a t i o n o f t h e Appendix: (3) Jt J i s always p o s i t i v e - d e f i n i t e . The where p o s s i b i l i t i e s o f approaching t h e minimum are then R , which contains dependent, mainly, on t h e m a t r i x t h e r e s i d u a l s and t h e i r second d e r i v a t i v e s . Obviously, and according t o t h e preceding e x p o s i t i o n , i f we succeed by means o f an a u x i l i a r y technique i n g e t t i n g a l l t h e elements o f R o f small magnitude, t h e a l g o r i t h m w i l l s u r e l y converge t o a minimum. T h i s i s p r e c i s e l y t h e reason f o r t h e e f f i c i e n c y o f t h e technique proposed i n Ref. [ l a ] , i n which t h e i n i t i a l process o f approaching i s assigned t o a F i r s t Order Gradient Technique, where t h e r e s u l t s are, as i n d i c a t e d , very e f f e c t i v e . 0. = On t h e o t h e r hand, as VF(Xp) f o l l o w s t h a t , a t t h e minimum where VF(Xp) then 9 must be t h e minimum, t h e accurate b u t t h e To a v o i d t h i s , must be used. 2, .Ut D it if D # @ s i n g u l a r . Consequently, as we approach successive steps become more and more Jacobian M a t r i x i s more d e t e r i o r a t e d . Jackson’s s i n g u l a r decomposition [31 GENERALIZED INVERSE METHOD T h i s technique i s presented i n [ll] and [13] applied to the parameter estimation using as mathematical model t h a t corresponding t o t h e Resistance measurements through t h e Driven Rod Method. As a complement i t i s i n d i c a t e d i n [13] t h a t t h i s technique can be used when t h e measurement method i s t h e Wenner Method. I n such case parameters pl, p 2 , z are used. I n our case we have examined t h e p o s s i b i l i t i e s o f t h e Generalized I n v e r s e Method o b t a i n i n g s y s t e m a t i c a l l y t h e f i e l d measurements through t h e Wenner Method and u s i n g as parameters pl, U, z. The o u t l i n e o f t h e method i s as f o l l o w s : L e t us assume t h a t we. are a b l e t o estimate a s e t o f parameters X . (i = l...v, i n our case v = 31, such t h a t t h e r e s i s t i v i t y values obtained through t h e r. w i l l be mathematical model f o r every spacing J s u f f i c i e n t l y c l o s e t o t h e measured r e s i s t i v i t y values p m ( r . ) , so t h a t we can evaluate them from an T a y l o r J expansion around t h e computed values p (X,r.) a - J That i s : The weighted d i f f e r e n c e ( e r r o r ) between measured and computed values w i l l appear i n t h e f o l l o w i n g form f o r every spacing r . J LEVENBERG-MARQUARDT METHOD The aim o f t h e Levenberg Method i s t o search f o r , a t each s t e p o f approaching t h e minimum, t h e smallest value reached by t h e preceding approximation of F(_X), i n such a way t h a t t h e change AX r e q u i r e d t o f i n d i t i s r e s t r i c t e d t o a hypersphere o f u n i t r a d i u s which c e n t e r s on XP. I n o t h e r words, a t each step we t r y t o : minimize such t h a t Making the F(X) AXt. derivatives AX of = 1 the corresponding 111 j = l...m J ; i i = 1...3 and t a k i n g i n t o account t h a t t h e expressions o f t h e elements o f D and J , (see l i s t o f symbols and Appendix), i t f o l l o w s t h a t f o r t h e v e c t o r E w i t h m elements e . : J E = D + Jt AX We e s t a b l i s h t h a t t h e best e s t i m a t i o n o f t h e parameters o f t h e model w i l l be t h a t which makes t h e Euclidean Norm square o f t h i s e r r o r v e c t o r minimum, that is: 1893 minimize 11 b).-Newton Method 112 E AXp f o r which = - [ Jt J c).-Levenberg-Marquardt a - 11 ax E 112 = 0 that is, for j = 1. . .m = - Axp d).-Generalized 2 Jt D and t h e second 2 ( Jt J 1, R I-' .Jt D [ Jt J + n I I-' Jt D I n v e r s e Method AXp Consulting t h e Appendix we see t h a t t h e f i r s t term i n t h e preceding expression i s : + Method f; - [ lIt 9 e).-Quasi-Newton Method XP+l - Xp = A Xp = - l7-l I-' Jt D ( V F(Xp) f o r m u l a t i o n necessary methods has been presented. All to - V F(Xp+')) implement such t h e r e f o r e , t h e sequence o f minimum searching i s : IMPLEMENTATION AXP - [ .Ut J I-'Jt D (5) Note t h a t i n t h i s case, t h e f u n c t i o n t o be optimized i s n o t t h e Euclidean Norm square o f D, as i n t h e preceding techniques, b u t t h e Euclidean Norm square o f E. I n any case t h e f i t i s performed over weighted d i f f e r e n c e s . For comparative purposes, when a p p l y i n g t h e Generalized I n v e r s e Technique, we w i l l a l s o The techniques implemented so f a r , i n the parameter e s t i m a t i o n problem o f a ground shaped as two-layers, from measurements made u s i n g Wenner's method have been based on: F i r s t Order Gradient [ 9 ] [ 6 ] [ l o ] , Generalized I n v e r s e [13] and Second Order Gradient [18]. These techniques are implemented here i n order t o compare them w i t h t h e new proposed techniques, and are a p p l i e d t o t h e same cases. evaluate t h e o b j e c t i v e f u n c t i o n 11 D 112 The customary d i f f i c u l t i e s i n t h e a p p l i c a t i o n o f this technique are similar to those of the Levenberg-Marquardt Method, as r e l a t e d t o t h e product Jt J . Consequently t h e Jackson S i n g u l a r Decomposition [3] should be applied. Indeed, if JtJ is approximately s i n g u l a r , one o r more o f i t s eigenvalues c o u l d be c l o s e t o zero. I t can be seen t h a t a small eigenvalue w i l l cause a l a r g e change i n one o r more o f t h e elements o f A5. Similarly, given t h a t t h e variances o f t h e model parameters are i n v e r s e l y p r o p o r t i o n a l t o t h e square r o o t o f t h e eigenvalues [ 3 ] , i t may happen t h a t small values o f variance w i l l y i e l d l a r g e standard d e v i a t i o n s . QUASI-NEWTON METHOD This term refers t o the methods i n which t h e minimum searching d i r e c t i o n a t every p o i n t t h e form - A VF(Xp), where A Xp i s of i s a positive-definite m a t r i x approximating t h e i n v e r s e o f t h e Hessian M a t r i x . The gradient direction is then deflected by p r e m u l t i p l y i n g i t by A . Probably, t h e Quasi-Newton Methods a r e t h e most e f f e c t i v e n o n l i n e a r o p t i m i z a t i o n methods f o r general problems [14]. I n a Quasi-Newton Method, the d i r e c t i o n o f movement towards t h e minimum i s e s t a b l i s h e d such t h a t : The a l g o r i t h m used f o r t h e treatment o f t h e problem we are c o n s i d e r i n g i n t h e a n a l y s i s o f t h e cases proposed i n 1181, i s t h e same one proposed i n 141, and now implemented by means o f t h e subroutine ZXMIN from t h e IMSL l i b r a r y [81. GENERAL CONSIDERATIONS For each implemented technique: -The s t a r t i n g p o i n t i s from a s e t o f f i e l d r e s i s t i v i t y measurements, u s i n g Wenner's Method. -The i n i t i a l values f o r each i t e r a t i v e process are: For py , t h e r e s i s t i v i t y corresponding t o t h e smaller d i s t a n c e o f t h e measurements l i s t For , pi t h e r e s i s t i v i t y corresponding t o t h e g r e a t e r d i s t a n c e o f t h e measurements l i s t . KO, i f a p p l i c a b l e , For t h e preceding c r i t e r i a . For zo, a concordant value w i t h 1 m i n a l l cases. - A l l t h e software Programs have been executed i n a HP, Vectra RS 20/C w i t h mathematical coprocessor. For t h e purposes o f f i n a l comparison, t h e value o f t h e same O b j e c t i v e Function F, t h a t i s , 1) D 112, t h e Euclidean Norm o f OF, t h a t i s )(VF(X)II, and t h e run t i m e are evaluated. CASES TO BE SOLVED SUMMARY The r e s u l t s i n r e l a t i o n t o i t e r a t i v e process f o r m u l a t i o n i n t h e search f o r t h e Optimum i n t h e d i f f e r e n t methods examined f o r t h e treatment o f our problem, can be summarized as: (see l i s t o f symbols) a).-Steepest I n a d d i t i o n t o t h e preceding techniques, t h e f o l l o w i n g techniques are a p p l i e d t o s i x s e l e c t e d t e s t cases [18]: An improved F i r s t Order Gradient which a l l o w s reaching t h e absolute minimum o f t h e o b j e c t i v e f u n c t i o n ; a technique based on t h e Levenberg-Marquardt method; another one based on t h e Generalized I n v e r s e Method s i m i l a r t o [ I 3 1 b u t a p p l i e d t o d i f f e r e n t s e t o f parameters; one more based on a Quasi-Newton Method and; f i n a l l y , a mixed Second Order-Generalized I n v e r s e technique which a l l o w s an improvement over t h e one proposed i n [18] regarding run-time. Descent method AXp = - I-( Jt D The techniques w i l l be a p p l i e d t o s i x cases proposed i n [18]. Case U1 corresponds t o references [61 [lo]. Most o f t h e Programs have been b u i l t on TurboC and TurboC++ (Borland) v e r s i o n 2.0. Two o f t h e techniques have been implemented u s i n g IMSL r o u t i n e s , i n i t i a l l y foreseen t o be run on VAX 11 and adapted t o be compiled i n FORTRAN M i c r o s o f t Version 4.0 and L i n k e r Version 5.01.20. 1894 CAHRT. METHODS OF [hl AND Fig.l.-FLOW [e] Meas: File *.OAT = Niterup /Guess a vector as s t a r t i n g p o i n t I I 500 COmPute I1 Compute Gradient's Components vn~(xP)= Normalize it 1 VF(&P)/lVF(Xp)~ F(~P: FOi??l THE CHANGES - = VF(Xp) - 0, I r 111 111 VF(x_C) normalize i t XVnFt:P) searching the maximum decrease o f F(z) I CORRECT the Vector t o the IV NEW value [ A I vector Build $-I = E~ + I V VI Ye;bSOLVE VII Fig.2.-FLOW CHART. PROPOSED FIRST ORDER GRADIENT TECHNIQUE Fig.4.-FLOW W R T . P R O W S E O TECHNIQUE BASED IN A GENERALIZED INVERSE METHOD Meas.Files *.DAT .Niterup = 500 R a n d m steps=RST=100 Heas: File *.DAT Niterup + ___) I1 C m p u t e Objective Funct. F(Xp) c m p u t e Gradient's c a n w n e n t s Normalize it VnF(Xp) I I k = + , I ' I1 ~~ I = 50 VF(X~) = VF(Xp)/lVF(~')~ I 11 I1 I11 I Generate r a n d m l y Build [XI vector Compute 5; = 5' -[XIhnF(xp) I C m p u t e D and I cmpute [ Jt J I-' Jt D i-' p:pt1 I Cmpute IV II "I1 ,m,o A cl Resu 1ts I t xP1z 3' - [JtJ]-'JtD ill Results 1 1895 1 s t .-IMPLEMENTATION OF THE METHOD PROPOSED I N ANSI/IEEE Std.81-1983, Ref. 191 T h i s i s a F i r s t Order Gradient Technique (FOGT). It uses as parameters pl, p2 , z, and i t has been implemented according t o t h e Flow Chart shown i n F i g . l , by means o f a TurboC Program. I n t h e Stage 11, f o r t h e c a l c u l a t i o n o f t h e O b j e c t i v e Function and f o r lack o f an e x p l i c i t c r i t e r i o n i n 191, t h e r e q u i r e d terms a r e used u n t i l t h e new term brought t o t h e s e r i e s w i l l c o n t r i b u t e with a value l e s s than l.E-5 o f t h e sum c a r r i e d out. I n any case, t h e maximum number o f terms o f t h e s e r i e s i s r e s t r i c t e d t o 10,000. I n Stage 111, although n o t i n d i c a t e d i n Ref. [ 9 ] , we have normalized t h e Gradient, u s i n g f o r t h i s t h e Euclidean Norm, as otherwise i t i s impossible t o reach a solution. The elements o f t h e diagonal o f [XI have been taken as z = .005 Ipl I , .005 lp21 and I z I , (see Ref. [91). The c r i t e r i o n t o s t o p t h e i t e r a t i v e process when IAF(X)Itl.E-3, as i n d i c a t e d i n [91, has been v e r i f i e d as a b s o l u t e l y i n e f f e c t i v e , s i n c e t h e process f u l f i l l s t h i s c o n d i c t i o n almost immediately, s t i l l being f a r from t h e minimum, as can e a s i l y be checked. A value o f l.E-5 has been used instead. 2nd.-IMPLEMENTATION OF THE METHOD PROPOSED I N EPRI -Research P r o j e c t 1491 Report EL 2699, Ref. [61 and [ l o 1 T h i s i s a F i r s t Order Gradient Technique (FOGT). I t uses as parameters p , K , z. I t has been implemented u s i n g t h e same Flow Chart o f F i g . l by means o f a Program w r i t t e n i n TurboC. I n Stage 11, f o r t h e c a l c u l a t i o n o f t h e O b j e c t i v e Function, t h e r e q u i r e d terms a r e used u n t i l t h e new term brought t o t h e s e r i e s w i l l c o n t r i b u t e w i t h a value l e s s than .001 o f t h e sum c a r r i e d out. I n Stage 111, although t h i s i s n o t i n d i c a t e d i n Ref. [61, we have normalized t h e Gradient, u s i n g f o r t h i s t h e Euclidean Norm, as suggested i n , as otherwise i t i s impossible t o reach a s o l u t i o n . I n Refs.[6] and [ l o ] no c r i t e r i o n i s s p e c i f i e d f o r t h e s e l e c t i o n o f [1], except t h a t a "proper s e l e c t i o n " must be made. Thus we assume t h a t t h e authors adopt f o r t h e diagonal m a t r i x [ X I a value according t o R e f . [ 9 ] , t h a t i s T; = .005 IplI , .005 I P 2 1 and .005 I z I . The c r i t e r i o n t o s t o p t h e i t e r a t i v e process when IAF(X)I<l.E-3, as i n d i c a t e d i n [61 1101 has been v e r i f i e d as a b s o l u t e l y i n e f f e c t i v e , s i n c e t h e process f u l f i l l s t h i s c o n d i t i o n almost immediately, s t i l l being f a r from t h e minimum, as can e a s i l y be checked. A value o f l.E-5 has been used instead. 3rd.-PROPOSAL OF TECHNIQUE is AN IMPROVED FIRST ORDER GRADIENT An improved F i r s t Order Gradient Technique (FOGT) presented, u s i n g as parameters K, Z, implemented f o l l o w i n g t h e f l o w diagram o f Fig. means o f a TurboC Program. 2, by terms of t h e s e r i e s i s r e s t r i c t e d t o 1,000. I n Stage I V , a random p o s i t i v e number i n i n t e r v a l ( 0 , l ) i s generated f o r t h e s c a l e f a c t o r T h i s i s s i m i l a r t o t h e treatment t h a t used i n [ l a ] one o f i t s c a l c u l a t i o n stages. We use as v e c t o r scale f a c t o r s 1000 1. .02 1 and .75 1 the 1. in of The Euclidean Norm o f t h e Gradient v e c t o r v e r i f i e d t o be l e s s than l.E-3 t o make a p r i n t - o u t the results. is of 4th.-SECOND ORDER GRADIENT TECHNIQUE (SOGT) I t i s a technique based on t h e Newton Method which takes advantage, o f t h e behavior o f a F i r s t Order Gradient technique f o r approaching t h e minimum, and o f t.he q u a d r a t i c convergence o f t h e Newton Method when i t i s i n an area c l o s e t o t h e s o l u t i o n . The a l g o r i t h m was described i n Ref. implemented i n TurboC. I t s Flow Chart i s Fig.3. 5th.-TECHNIQUE (LMT 1 BASED ON We use as parameters [181 and shown i n LEVENBERG-MARQUARDT METHOD , z. p, K I t has been implemented by means o f t h e Routine ZXSSQ from Ref. [81 w i t h o u t e x p l i c i t c a l c u l a t i o n o f t h e f i r s t d e r i v a t i v e s . For c a l c u l a t e d apparent r e s i s t i v i t y , t h e I n t e g r a l Form [181 has been used, given t h a t t h e program i s being used t o i n v e s t i g a t e t h e p o s s i b i l i t i e s o f m u l t i - l a y e r modelling. The r e s u l t s do n o t d i f f e r from those obtained u s i n g t h e i n f i n i t e s e r i e s . The source code used t o d e f i n e t h e Function has been implemented i n FORTRAN 77. For t h e e v a l u a t i o n o f t h e i n t e g r a l , t h e Romberg Method has been used, being implemented [17] by means o f QROMO, MIDINF and POLINT Routines. The BesSel Function values have been approximated by polynomials using the Function BESSJO(X) from Ref. 1171. As convergence c r i t e r i o n , t h e Euclidean Norm o f t h e Gradient o f t h e O b j e c t i v e Function i s r e q u i r e d t o be smaller than l.E-5. 6th. -TECHNIQUE BASED ON THE INVERSE GENERALIZED METHOD (IGT 1 We use as parameters pl, , K z. I t has been implemented using t h e Flow Chart shown i n Fig.4, by means o f a TurboC Program. The process i s stopped when t h e v a r i a t i o n i n each parameter i s l e s s than l.E-5. The Program has been implemented i n such a way as t o r e j e c t t h e measurements o u t s i d e o f an a d j u s t a b l e range o f e r r o r , by d e f a u l t 15%, t o avoid s i n g u l a r i t i e s . 7th.-TECHNIQUE BASED ON A QUASI-NEWTON METHOD (QNT) We use as parameters pl, K , z. I t has been implemented by means o f ZXMIN Routine from R e f . [ 8 ] , w i t h o u t e x p l i c i t c a l c u l a t i o n o f t h e f i r s t d e r i v a t i v e s . For c a l c u l a t e d apparent r e s i s t i v i t y t h e I n t e g r a l Form has been used, because t h e program i s being used t o i n v e s t i g a t e the possibilities o f multi-layer modelling. The Hessian Matrix is i n i t i a l i z e d as a u n i t m a t r i x . I n Stage 11, t h e sum o f t h e terms f o r t h e c a l c u l a t i o n o f F(8) i s stopped when t h e c o n t r i b u t i o n o f 8th.-TECHNIQUE a new term t o t h e s e r i e s i s l e s s than 1.E-5 o f t h e sum already c a r r i e d out. I n any case, t h e maximum number o f The exposed considerations concerning the convergence c h a r a c t e r i s t i c s o f t h e Newton Method when BASED ON A MIXED METHOD (MMT) 1896 t h e process of approaching t h e minimum i s c a r r i e d o u t i n t h e p r o x i m i t i e s o f t h e minimum, were already used favorably i n [181, s t a t i n g t h e t o t a l e f f e c t i v e n e s s o f t h e procedure. acceptable values i n almost any cases. The lst(F0GT) Technique achieves acceptable values, which does not happen with t h e 2th(FOGT). TABLE V I I 1 . - M I X E D I t i s expected t h a t , improving t h e i n i t i a l values, t h e process w i l l converge w i t h g r e a t e r s e c u r i t y and rapidity to the minimum. Among t h e techniques examinated, as we w i l l see i n t h e f o l l o w i n g paragraph, t h e one which reaches t o r e s u l t s c l o s e t o t h e optimum w i t h a h i g h r a p i d i t y i s t h e technique based on t h e I G T , and w i t h enough accuracy t o be u s e f u l i n a h i g h number o f practical applications. METHOD RESULTS: Iterations : 8 o f INVERSE GENERALIZED TECHNIQUE (0.08 min) 1 FIRST ORDER GRADIENT and 4 SECOND ORDER GRADIENT (0.40 min) GRADIENT Maximum p e r m i s s i b l e value 1E-005 The proposal c o n s i s t s i n u s i n g as i n i t i a l values, f o r t h e Newton Technique presented i n [18], those p r e v i o u s l y generated by t h e technique based on t h e I G T as presented i n t h i s paper. To o b t a i n t h e Flow Chart, i t i s s u f f i c i e n t t o l i n k those o f F i g ' s 4 and 3. HESSIAN M a t r i x . Last i t e r a t i o n RESULTS STARTING VALUES (two-layer s t r u c t u r e ) UPPER LAYER R e s i s t i v i t y (0.m) = 30.2 BOTTOM LAYER R e s i s t i v i t y (0.m) = 7.1 DEPTH UPPER l a y e r (m) = 1 OBJECTIVE FUNCTION = 6.35825484 I n Tables I through V I , t h e values obtained by a p p l i c a t i o n of these techniques t o t h e proposed cases i n [181 are c o l l e c t e d . The estimated values f o r pl, p2, z, t h e value o f t h e O b j e c t i v e Function F, Euclidean Norm o f t h e Gradient and Run Time a r e reported. The e x c e l l e n t r e s u l t s obtained can be appreciated, i n general, with t h e 4th(SOGT), 5th(LMT), 7th(QNT) and 8th(MMT) techniques. I n Table V I 1 t h e number o f i t e r a t i o n s r e q u i r e d f o r t h e MMT, compared t o t h e SOGT i s shown, where i t can be appreciated t h a t t h e MMT i s even b e t t e r than t h e SOGT from Ref. [181. T h i s i s t r u e w i t h respect t h e run-times, which a r e d r a s t i c a l l y smaller than those of t h e SOGT. I t must be noted, on t h e o t h e r hand, t h a t i n t h e proposed MMT, t h e i t e r a t i o n s o f F i r s t Order Gradient Technique a r e almost unnecessary. I t i s v e r i f i e d , i n d i r e c t l y , t h a t i n t h e process o f approaching t h e minimum, t h e I G T i s much more e f f e c t i v e t h a t any FOGT 1, as i n t h e case which employs t h e diagonal m a t r i x o f t h e SOGT when i t uses a FOGT i n i t s process o f approaching t h e minimum. (SOGT) Ref.[91 1 s t Order 2nd Order (MMT) 1 s t Order This paper 2nd Order I I 14 2 4 27 5 11 9 10 8 7 9 21 1 0 1 2 0 4 5 5 5 5 5 6 I As an example, t h e corresponding r e s u l t s , from a p p l y i n g t h i s MMT technique, t o a s e t o f measured values f o r i m p l a n t a t i o n o f a Substation i n Meleno, LeganBs, MADRID (Spain), a r e l i s t e d i n Table V I I I . The discrepancy i s expressed i n percent as: 100 (pm-pa)/pa I n r e l a t i o n w i t h t h e value o f t h e O b j e c t i v e f u n c t i o n F. t h e 4 t h (SOGT), 8 t h (MMT), 5 t h (LMT), 7 t h (QNT), and 3 r d (FOGT) Techniques are p r a c t i c a l l y e q u i v a l e n t , where indeed t h e F i r s t Order Gradient technique, proposed i n 3 r d p l a c e i s included. The 6 t h (IGMT), 1 s t (FOGT) and 2nd (FOGT) techniques , i n t h i s order have been shown t o be s l i g h t l y i n f e r i o r . From t h e p o i n t o f view o f Euclidean Norm o f t h e Gradient, o n l y t h e 4 t h (SOGT), 8 t h (MMT), 5th(LMT), 7th(QNT) and 3rd(FOGT) Techniques reach a s u f f i c i e n t l y small value. The Gth(1GT) Technique does n o t g i v e 0.0299 0.0273 0.3353 0.0273 0.1041 1.9913 0.3353 1.9913 56.5607 COMPUTED VALUES (two-laver s t r u c t u r e ) 29: 800 UPPER LAYER R e s i s t i v i t y - (0.m) = BOTTOM LAYER R e s i s t i v i t y ( 0 . m ) = 5.635 DEPTH UPPER l a y e r (m) 10.238 OBJECTIVE FUNCTION (end-value) Euclidean Norm o f t h e GRADIENT Meas.# Spacing(m) 1 2 3 4 5 6 7 1.500 2.000 2.500 3.000 3.500 4.000 4.500 5.000 8 9 10 11 12 13 14 15 16 17 18 5.500 6.000 8.000 10.000 13.000 16.000 20.000 25.000 32.000 40.000 p,meas. 30.200 29.700 30.000 29.000 28.900 28.800 28.800 27.600 27.400 28.200 27.000 24.000 18.900 17.500 13.000 10.000 7.800 7.100 = 0.01387167 3.44657E-009 PaComP. 29.756 29.698 29.605 29.472 29.294 29.068 28.794 28.473 28.105 27.694 25.693 23.340 19.722 16.481 13.085 10.235 8.030 6.864 d i screp. x; 1.491 0.006 1.333 -1.601 -1.344 -0.923 0.019 -3.065 -2.508 1.828 5.088 2.826 -4.167 6.185 -0.651 -2.299 -2.868 3.445 I n r e l a t i o n w i t h t h e t i m e needed t o o b t a i n t h e r e s u l t s . i t i s obvious t h a t t h e a u i c k e s t o f a l l t h e techniques i s t h e 6 t h (IGT), followed, i n most cases, by t h e 8 t h (MMT). The 4th(SOGT) and 5 t h (LMT) techniques present i n general a r e l a t i v e l y reduced run-time. The 3rd(FOGT) and 7th(QNT) techniques present g e n e r a l l y a p r o h i b i t i v e time, which i s excusable o n l y i n t h e second because o f t h e h i g h accuracy obtained. The lst(F0GT) and 2nd(FOGT) techniques present, normally, s i m i l a r and r e l a t i v e l y acceptable run-times. ~~ CONCLUSIONS The t h e o r e t i c a l b a s i s f o r t h e e s t i m a t i o n o f a vector o f parameters, o b t a i n i n g an optimum f i t t i n g between a s e t o f measured r e s i s t i v i t y values and t h e corresponding s e t o f computed r e s i s t i v i t y values u s i n g such parameters has been reviewed. Our f i r s t o b j e c t i v e has been t o d e f i n e t h e necessary t o o l s f o r l a t e r implementation o f e i g h t d i f f e r e n t techniques which w i l l a l l o w t h e e s t i m a t i o n o f parameters o f a two-layer earth. 1 Used Used COMPUTED VALUES Method p1 z p2 F UV FI 1s t ------b i n > (min) 1s t 355.773 143.122 2.880 .008361 2.23E-3 ----324.600 141.454 3.176 .014998 4.18E-2 ----360.379 143.751 2.821 .007950 9.04E-4 ----- 2nd 3r d 4th 6th 3r d .00162 7.32E-8 4th 372.727 145.262 2.689 .007627 1.85E-5 5th 364.683 143.634 2.828 .00811 -l-l-l 5th 2nd 144.470 2.760 368.295 0.05023 0.06 7th 8th 0.58 TABLE I 1 RESULTS i n CASE I 2 Used z P2 p1 10 Fi Runtime 13.72 491.026 92.92 .0112 4.431 5.03E-8 4 th 5th I 241.2821-I -I 1 I 6th I 992.25~1.950~.011052~.047131 0.08 I1-1246.836~1058.62~2.139~.010740~9.01E-93.12 7th I I 7th 8th I 246.836~1058.6312.139~.010740~8.46E-8I 0.55 I 8th T E G l G l G l ~ l ~ l ~ RESULTS i n CASE t 5 RESULTS i n CASE t 3 -----170.582 48.896 1.452 .013698 4.14E-3 0.617 -----158.920 35.323 1.829 .019508 2.98E-2 0.38 -----161.187 1. 34.802 160.774 ~ 167.252 34.068 l 45.306 RV F[ (m) lI ~ 1.570 .012834 .lo8191 p2 (0.m) F (min) ----123.941 5 6 . 8 5 1 9 4 . 4 8 9 ~ 1 . 5 3 6 ~ . 0 2 4 2 0 2 ~ 4 . 2 5 E - 3 0.33 2nd 132.934 1055.24 2.821 .022387 1.27E-1 3r d 125.474 1 0 9 2 . 5 j 2.71~.0207Oj8.76E-4 4th 122.319 1035.78 2.465 5th 5th 125.527 1092.831 2.712 6th 6th 117.952 7th 125.525 8th 125.526 3 rd 4th 7th 8th l~lGlGl~lGl;l 57.279 96.489 1.626 .024027 7.33E-4 l*-l%GlLiElzl 57.344 L 96.714 1.651 .024018 4.04E-9 I I I I I J 0.45 FI Runtime (m) 1st 2nd 0.04 CORPUTED VALUES (min) 1 1 ~ 0.97 RESULTS I n CASE #6 Method Runtime 1.848 .010782 7.27E-6 160.7761 34.074(1.8481.010383(1.52E-8 Used F 1 6.98 164.35 39.88 /1.6981.0107 11.52E-7 1.35 ------ TABLE V I COMPUTED VALUES 1.828 .010897 7.01E-4 57.139 93.243 1.517 .024314 4.OE-3 0.36 ------ 1st 3.31 Method 5th 1 .011207 8.22E-4 Used F 3r d I 93.395 4.399 494.883 4th p2 (0.m) 490.688 8th 2nd p1 (0.m) 1.63 494.883 3 rd Used 4.13 .055909 1.23E-1 7th 2nd Method -1-1-1-1-1- I 100.403 4.081 468.897 1s t TABLE I11 92.27614.551(.01385316.18E-3 463.851 6th 1s t 6th 470.0181 TABLE V COMPUTED VALUES Method I CMPUTED VALUES Method Runtime (0.m) (0.m) (m) ----- 1897 RESULTS i n CASE I 4 TABLE I V 994.3612.534) .02130314.53E-3 ----- I ,0206 4.60E-6 ----.020702 Il.06E-4 I 4.47 I I 1 1 2.03 4.70 2.25 1.30 l 1898 As i s t y p i c a l i n t h e Steepest Descent Method i n t h e d i r e c t i o n o f t h e Negative Gradient, t h e r e i s an important s c a l e f a c t o r t o be taken i n t o account. An improper s e l e c t i o n o f [ a i may cause a precarious convergence o f t h e process. The r e s u l t s o f t h e s i x cases examined, and many o t h e r s which cannot be presented here because o f l a c k o f space, show a s a t i s f a c t o r y behavior o f t h e Technique from Ref. 191 w i t h t h e introduced m o d i f i c a t i o n s . A s t i l l b e t t e r behavior i s presented by t h e FOGT proposed i n t h i s paper which p r o f i t s , as i n d i c a t e d , from p a r t o f t h e philosophy shown i n 1181, t o generate t h e diagonal matrix [ X I . ACKNOWLEDGEMENT As indicated, the results and conclusions presented i n t h i s paper a r e a p a r t of a Research P r o j e c t which attempts t o d e f i n e t h e b e s t t o o l f o r e s t i m a t i o n o f s o i l parameters i n a m u l t i - l a y e r e a r t h . The Author wishes t o express h i s g r a t i t u d e b o t h t o OCIDE and IBERDROLA f o r sponsoring t h i s P r o j e c t . He would a l s o l i k e t o p o i n t o u t t h e support received from t h e Engineers o f t h e D I D I S Department o f IBERDROLA i n t h e development o f t h i s P r o j e c t . REFERENCES The FOGT proposed i s , i n a l l t h e cases examined, b e t t e r t h a n those from Ref. [61 [91 [ l o ] , w i t h repect t o a b e t t e r f i t n e s s o f values. However, t h e execution t i m e i s , i n a l l cases, higher. not #4). [91. are, The 2nd Technique (FOGT) o f t h e Ref’s [ 6 ] [ l o ] i s v e r y e f f e c t i v e i n some cases ( f o r example #1 and I n every case i t i s i n f e r i o r t o t h a t from r e f . The values o f t h e components o f f i n a l Gradient t h e m a j o r i t y o f t h e times, f a r from zero. The 1 s t Technique (FOGT) proposed i n Ref.[9], w i t h t h e m o d i f i c a t i o n s introduced i n t h i s paper, leads t o s a t i s f a c t o r y values and can be used i n t h e parameter e s t i m a t i o n o f a two-layer ground. I n any case, none o f t h e FOGT analyzed presents a s u f f i c i e n t r e l i a b l e behavior i n t h e searching f o r t h e optimum adjustment under any circumstance, f o r i t t o be taken as a b a s i s f o r t h e optimum parameter e s t i m a t i o n i n multi-layer earth. The 6 t h Technique (IGT) r e s u l t presents components o f t h e g r a d i e n t t o o h i g h t o perform a s u f f i c i e n t approximation t o t h e optimum. Nevertheless i t can be u s e f u l i n p r a c t i c e due t o t h e r a p i d i t y i n o b t a i n i n g r e s u l t s . I n any case, i t can be considered, as has be done i n t h e MMT, as a very useful a u x i l i a r y t o o l f o r approaching t h e minimum f o r a l a t e r use i n a more e f f e c t i v e Technique. Compared t o t h e r e s t o f Techniques, t h e 8th(MMT) proposed presents an e x c e l l e n t behavior, a c h i e v i n g i n most cases t h e smallest values o f F(X), uVF(X)II, i n t h e same way as t h e 4thCSOGT). Furthermore t h e lowest execution times a r e achieved. S i m i l a r c h a r a c t e r i s t i c s are provided by t h e Sth(LMT1, though w i t h execution t i m e always s l i g h t l y higher. The time f o r t h e 7th(QNM) i s a l s o very high, b u t i t s accuracy i s o f t h e same order, even sometwhat b e t t e r than t h a t from 8th(MMT), Ith(S0GT) and 5th(LMT) Techniques. From t h e preceding e x p o s i t i o n i t i s concluded t h a t t h e recommendations f o r e s t i m a t i n g t h e parameters o f a two-layer earth, and in decreasing order of e f f e c t i v e n e s s , g i v i n g weight t o accuracy ( t h e most important) and computer run time, are: Mixed Technique Second Order-Generalized I n v e r s e (MMT), Technique based on t h e Levenberg-Marquardt (LMT), Technique based on Newton Method (SOGT), Technique based on a Quasi-Newton Method (QNT), proposed 3 r d Technique o f F i r s t Order Gradient technique (FOGT), F i r s t Order Gradient [ l o ] and f i n a l l y t h e I n v e r s e Technique from [ 6 ] Generalized Technique (IGT). The t h r e e f i r s t c l a s s i f i e d present c h a r a c t e r i s t i c s o f accuracy and run-time t h a t convert them i n t o powerful a n a l y s i s t o o l s f o r t h e s o l u t i o n t o t h e problem o f parameter e s t i m a t i o n i n two-layer e a r t h . The I G T must be used as a u x i l i a r y f o r t h e generation o f i n i t i a l values. E.D., Earth conduction effects i n ill.-Sunde Transmission Systems. Dover P u b l i c a t i o n s . (1968) [ 2 l . - F l e t c h e r R., A M o d i f i e d Marquardt subroutine f o r non-1 i n e a r l e a s t squares. Harwell, Berkshire, England. Atomic Energy Research Establishment. Report No. AERE-R.6799 (1971) D.D., Interpretation of inaccurate, [3l.-Jackson i n s u f f i c i e n t and i n c o n s i s t e n t data. Geophys J.R. A s t r . SOC. 28, pp 97-109 (1972) [41 .-Fletcher R., FORTRAN subroutines f o r m i n i m i z a t i o n by quasi-Newton methods. Harwell, Berkshire, England. Atomic Energy Research Establishment. Report No. AERE-R.7125 (1972) [5].-Bazaraa M.S., Shetty C.M., Non L i n e a r Programming. Theory and Algorithms. Ed. John Wiley L Sons. (1979) [6].-Electric Power Research I n s t i t u t e , Transmission L i n e Grounding. Vol 1. Research P r o j e c t 1494-1. Report EL-2699. P r i n c i p a l Author F.Dawalibi. (1982) [71.-Blatner C.J., Study of d r i v e n ground rods and f o u r p o i n t s o i l r e s i s t i v i t y t e s t . I E E E PAS. Vol 101, Aug (1982), pp 2837-2850 [8].-IMSL Routines, 9 t h e d i t i o n (FORTRAN). Houston, Texas. (1982) IMSL I n c . [9].-IEEE Std. 81., I E E E Guide f o r Measuring E a r t h Resistivity, Ground Impedance and E a r t h Surface P o t e n t i a l s o f a Ground System., (1983) [lO].-Dawalibi F., B l a t t n e r C.J., Earth r e s i s t i v i t y measurement i n t e r p r e t a t i o n techniques. I E E E PAS. Feb (19841, pp 374-382 [Ill.-Meliopoulos A.F., e t al., Estimation o f Parameters from D r i v e n Rod Measurements. PAS, No 9, (19841, pp 2579-2585 Soil IEEE [12].-Blatner C.J., Analysis o f s o i l r e s i s t i v i t y t e s t methods i n two-layer e a r t h . I E E E Transactions on PAS-104, No 12, (19851, pp 3603-3608 [13].-Meliopoulos A.P., Papalexopoulos A.D., Interpret a t i o n o f r e s i s t i v i t y measurements. Experience w i t h t h e model SOIMP. I E E E Transactions on Power D e l i v e r y , VOL PWRD 1, Oct (1986), pp 142-151 [14].-Cuthbert T.R., Optimization using computers. Ed. John Wiley L Sons (1987) [15].-Meliopoulos Transients. (1988). personal A.P.S., Power System Grounding and Ed. Marcel Dekker I n c . New York 1899 1161.-Takahashi T., Kawase T., Analysis o f Apparent R e s i s t i v i t y i n a Multi-Layer E a r t h S t r u c t u r e . I E E E Transactions on Power D e l i v e r y . Vol. 5, No. 2, pp 604-610 (1990) [17].-Press W.H., Flannery B.P., Teukolsky S.A., Vetterling W.T., Numerical Recipes (FORTRAN Version). Ed. Cambridge U n i v e r s i t y Press. (1990) A Second Order Gradient [18].-Alamo J.L. del, Technique f o r an Improved E s t i m a t i o n o f S o i l Parameters i n a Two-layer Earth. I E E E Transactions on Power D e l i v e r y . Vol. PWRD 6, No. 3, (19911, pp 1166-1 170 Expressions f o r aPa/aXi can be found i n [ 6 ] and [ 9 ] COMPONENTS OF THE HESSIAN MATRIX The Hessian M a t r i x components a r e r e s p e c t i v e l y : n = I...”, i = 1...3) (For any case j = l.......m, = 2 1 ( Aad] 2 + 2 1 - Ja24 d , axi2 ax j axi2 J APPENDIX GRADIENT COMPONENTS Remembering t h a t : F(X) represents t h e O b j e c t i v e Function F(p,,K,z), /I F ( P ~ , P ~ , Z ) .I n any case F ( X ) where t h e elements o f differences d . between J D D are measured or t 112 D .D, t h e weighted and computed values. pa represents the pa(~,,K,z,rj) or mathematical model either P , ( P ~ , P ~ , Z , ~ ~ ) .I n general 3. i s t h e number o f parameters For a l l cases j = l.....m, n -=21;---dj=-21aF ad. 1 m ’ axi = I n our case l...*m, v 3 each one o f t h e parameters, then: Ji ad. axi 1 aP P, axi - -a 2 .-! 1Jji Jjk + 2 4- G:, a2d, dJ. o r , i n m a t r i x form: H z 2 ( J t J + R ) i f k where R i s a 3x3 square m a t r i x containing weighted d i f f e r e n c e s and t h e i r second d e r i v a t i v e s . 1axi2 a2d J j d. J and Rik i f k = 1% j the d, axiaxk BIOGRAPHY j = - More d e t a i l e d expression f o r these can be found i n [18] d. If 9 represents t h e Jacobian M a t r i x o f p a r t i a l d e r i v a t i v e s o f t h e weighted d i f f e r e n c e s d j r e l a t e d t o J.. axi j axiaxk l1 aPa axi a2F -- R.. = i = 1. ..3 ad. Jgi+21J2dj 2- where pa(?, r j ) . v a2 F - - 2 1 ax r . i s the m~ J probe spacing a t which t h e measurements have been made, j = l . . . . m i s a b r i e f notation o f p ( r . ) , p, i f k and b e a r i n g i n mind t h e expression o f t h e elements Jji, it follows that: which i m p l i e s VF(X) = 2 .Ut The elements, i n a more e x p l i c i t form are: D J.L.de1 Alamo was born i n Albacete (Spain), on August 18, 1943. He received an E l e c t r i c a l Engineering degree from V a l l a d o l i d University (19631, a B.Sc. degree i n Phys. Sci. from S e v i l l a U n i v e r s i t y (19751, M.Sc. degree i n E l e c t r o n i c s from S e v i l l a U n i v e r s i t y (1978), and Ph.D. degree from V a l l a d o l i d U n i v e r s i t y (1984). From 1964 t o 1981 he worked f o r several Power System Companies i n S t a t i o n s , Substations and High Voltage Ingenieros Lines Design. I n 1981 he j o i n e d t h e E.T.S. I n d u s t r i a l e s , U n i v e r s i t y o f V a l l a d o l i d , and i s now F u l l Professor i n i t s E.E. Dept.