Scribd Logo
Upload

Welcome to Scribd!

  • Earthing With MATLAB

    Uploaded by

    Prakash Makhijani
    25 views6 pages
    Earthing desing using MATLAB

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document
    Earthing desing using MATLAB

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    Download as pdf
    25 views6 pages

    Earthing With MATLAB

    Uploaded by

    Prakash Makhijani
    Earthing desing using MATLAB

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    Download as pdf
    of 6
    2016 IE] E PES Asia-Pacific Power and Energy Conference - Xi'an — China Substation Grounding System Optimization with Utilizing a Novel MATLAB Application Xuan Wu Station Engineering Standards ‘American Electric Power Gahanna, OH, 43230 xwu@acpcom Abstract—the sim of this work is to develop an application ‘which has the capability of modeling and optimizing regular shape (rectangular, square or L. shape) ground grids under a tworlayer soil model. The ground grid optimal design is the focus of this paper by using a 3-step optimization method: 1) IEEE-80 equations to calculate touch and step potentials, Which makes constraints continuous and differentiable for sradient descent methods; 2) using the optimal solutions from step 1 as the initial input into a pattern search algorithm by applying a more accurate but non-differentiable touch and step potential calculation method; 3) performing a perturbation test to determine whether the results from step 2 are globally ‘optimal, otherwise using a genetic algorithm to re-optimize the solution from step 2 until it passes the perturbation test. Index Terms—ground. grid, optimization » calculation, software, touch potential 1. IntRopuction ODAY, more and more substations are created and reconstructed to satisfy the growing electricity demands for both industry and households, It is a big concem that the designed substation must guarantee the safety of persons who are in the area of the substation under both normal and fault conditions. As a result, an adequate grounding system is needed to protect people from touch and step potentials. Note that the touch potential is the only safety metric to justify if the designed grounding system is adequate in this paper. There are two methods to compute the touch potential, onc is an approximate method using experienced IEEE-80 equations Error! Reference source not found. and the other is an accurate calculation using numerical computations. Note that the calculated touch potential is needed to be compared with the allowable threshold for the safety assessment, Additionally, it is necessary to consider the cost savings when designing and constructing a grounding system. ‘Therefore, an optimization problem is proposed. Considering that each iteration in the optimization process using the more accurate numerical computations needs touch potential matrix calculations [2)+5]. which are disjunctive and _non- differentiable, it is impossible to utilize the traditional 978+1-5090-5417-6/16/$ 31.00 ©2016 IEEE Qianzhi School of Electrical Engineering University of Utah Salt Lake City, UT, 84112 gianzhi zhang@utah edu ‘hang Jiahong He School of Electrical Engineering Southeast University ‘Nanjing, China, 210000 ‘optimization method such as Newton method, Therefore, the heuristic method is the only option. Thus, the usage of heuristic methods is mainly discussed in this paper. Two case studies with different shapes (rectangular and L) Of substations are illustrated and corresponding MATLAB [6] application (OptimGrid) is introduced as well, Tl, Oprmaization StraTECy Presented in this section is a novel optimization strategy used in minimizing labor and material cost for installing a substation grounding system. The key requirement for an optimization method capable of —_grounding-system ‘optimization is that it must be able to deal with the discontinuous and non-differentiable constraints. The authors apply a new three-step optimization process combining a traditional optimization method with a heuristic probabilistic ‘optimization method, comprised of a pattem search method and genetic algorithms. A. Traditional Optimization Method using IEEE Equations The first step in the three-step procedure for designing an ‘optimal ground gird is to use approximate relationships, given by the IEEE-80 [1]. that are continuous and differentiable relationships to which traditional robust _ optimization techniques can be applied. The result of this optimization procedure will be the initial estimate for further utilization in the second step of the optimization process. In ground grid design, the mesh voltage is traditionally treated as the worst-case touch voltage on the earth surface above a grounding system and its value is obtained from following equation based on [1] ay where, = equivalent uniform soil resistivity (0m): Ku K, geometrical factor: corrective factor; os te A) Jy ~ effective grounding system conductor and rod total length (rm). maximum fault current through the grounding system ‘The detailed expressions and calculation process of p, Kn. Gy, Ig and Ly are deseribed and explained in [1]. While some of these parameters are exogenous parameters, ic. fp and p, the other parameters are functions of conductor spacing (D) in meters, depth of grounding grid (D.p) in meters, conductor diameter (dia) in meters, length (a) and width (6) of the substation in meters, mumber of rods (7) rod length (h) in meters, rod diameter (dia), fault current (J) in amps, and fault ceurrent division factor (19). Among them, only D and ne are the variables needed to minimize the objective as all other parameters are specified by SRP (Salt River Project, which was the sponsor ofthis researeh) as utility requirements, ‘As a result, equation (1) can be used in the constraint (6) in the Case Study. Since the objective function and constraints are all differentiable. a traditional optimization procedure can be used. Note that the L-shape grid is needed to be transferred to an equivalent square grid that has the same area with it. B. Pattern-Search Algorithm Patter Search method (PS) is a member of the family of direct search methods. A direct search algorithm searches a set of points around the current point, looking for one where the value of the objective function is Tower than the value at the ‘current point (in minimization problems). The most significant advantage of the direct search methods is that they do not require the objective function or constraint functions to be differentiable or continuous, Therefore, the more accurate non-differentiable and discontinuous constaints using matrix calculations can be applied in this second step. However, the disadvantage of direct search method and heuristic method (ike genetic algorithms discussed later) is the lengthy computational time. These methods are unlike sgadientsbased methods, which the user can implement ‘without knowing a direction to search for lower objective values. Instead, PS needs to test multiple points near the ccurrent point, just like a “mesh” around this point, which can be interpreted as: 1) Generating a set of vectors {01} by multiplying each pattem veetor v; by a scalar dy, which is called the mesh size. Content and generation of these vectors will be introduced with a numerical example presented later 2) Adding the {a} to the present point, which is the point with the best objective funetion value found at the previous step [7] For example, assume the present point has coordinates (3.6, 42] and its present mesh size (4) equals 8.0. In addition, the pattem vectors are set as: vr~ [1, OJ: v= [0. I]: vs~ [-1, 0} ve [0-1]. The vectors (di) are obtained as: d= 8.0*{1, 0] =[80, 0); d= 8.0*[0, 1] = [0, 80}; di 8.04[-1, 0] = [-80, 0]; d= 8.0%[0, -1] ~ [0, -80], The algorithm adds {ai} to the present point to obtain the following mesh: (3.6, 4.2}* [8.0, 0] = [115, 42); (36, 42}+ [0, 8.0] = (36, 122}; (3.6, 42}* [-80, 0} 54,42] [36.42]* [0,-8.0] [36, -38] If one of the four new points has the smallest objective among these five points, the new point will be set as the present point at the next iteration with a new mesh. However, if none of the new points has an objective function with a smallest value, the mesh size (4q) will be reduced. This mesh size reduction is continued until either a smaller value is found or the minimum size of dy is reached. C. Genetic Algorithm The genetic algorithm is the third (optional) step in the optimization procedure as well as the backup step [8]-{10} ‘This step is used if the pattem search approach does not yield the optimal design. The same constraint calculation method as in the second step is used. However, genetic algorithm has improved robustness (less reliable on the correct initial input) when compared with pattem search method, but it typically needs longer computational time. D__ Program Structure First, it is necessary to obtain the substation initial design data used as input data into this program. The initial data should include: 1) substation shape and dimensions; 2) ‘maximum fault current and division factor; 3) fault duration time; 4) soil model and parameters, which are obtained from @ developed soil modeling application (Appendix A); 5) resistivity and depth of the surface material, (If there is no hhigh resistivity surface material, the depth of surface material is treated as zero.) With the initial data, there are two touch potential calculation method, which are using IEEE approximate equations (ie. [1]} and numerical (matrix) computations [2]- [5]. A. benchmark between touch potential numerical ‘computations from OptimGrid and WinIGS [11] is shown in Appendix B. Note that the tolerable (allowable) touch potentials are calculated according to the formulas in [1} AAs stated earlier, the optimization process is a three-step procedure, in the first step, an approximate optimum is reached using the IEEE equation (1). In this application, the ‘optimization is achieved by using the “fmincon” solver in the MATLAB Optimization Toolbox as well as “GlobalSearch” function in MATLAB Global Optimization Toolbox to mode! this constrained nonlinear minimization problem [8]. With the help of the “GlobalSearch” function, the “fimineon” solver ean bbe made more efficiently and less dependently on the initial values. The details about the combination of “GlobalSearch” and “fimincon” ean be found in [6]. Although the optimization result is likely not as accurate as possible (though likely conservative) due to the approximations used in the derivations of that equation from [1], the result should not be too far from the globally optimal solution. Therefore. it can provide a good initial estimate of the optimal solution for the next step ofthis process. Im the second step of the optimization procedure, the ‘optimal solution from first step is used as an initial estimate for MATLAB's “patternsearch” solver. This will not only improve the efficiency of pattem search procedure, but also reduce the probability of ending with the locally optimal a9 solution, which is more likely to happen if a poor initial estimate is used, ‘After the pattem search procedure is exhausted, it is necessary to perform the perturbation test to determine ‘whether the point obtained is a locally optimal solution or not. For example, Fig. 1 shows a 2-by-2 ground grid with no ‘ground rods. It is assumed that this grid resolution is obtained after running pattem search. If this is the globally optimal solution, the constraints (6) will not be satisfied when making the grid resolution coarser. In other words, if itis possible to increase D until the grid resolution gets coarser (changed from ‘2-by-2 to I-by-l in tis example) and the objective function is reduced while the constraints are still satisfied, the optimal solution at (Ns, Nj, Ne) = 2, 2, 0) is not a globally optimal solution and the result is suboptimal. Note that the definitions Of Nc, Ny and Nie are introduced in Section III. Conversely. if the constraints are violated by making such a change, itis at least plausible that the result is locally (and possibly globally) optimal vn * Grounding «0 oa le 2» oo ae) Figue 1.24 Grounding Grid with X-¥ Coordinate Sytem Ifthe optimal result after running pattern search is. ( nd) = @, 3,4) in the example cited, ts necessary not only 0 verify the plausibility of the solution being optimal by changing (Ns, Ny, Nod) ~ @, 2.4) to (1,1, 4), which means ‘making the grid resolution coarser. In addition, one must also verify. the plausibility of the solution being optimal by changing 2,2, 4) to 2, 2,3), which means deleting one ro In addition to making the gid coarser, deleting, rod(s) is another way of making the objective function smaller, a change that would violate the constraints if the solution were ‘optimal After completing the perturbation test and finding that the solution is not optimal, as the third step in the optimization procedure is implementing the genetic algorithm. The genetic algorithm used here takes more execution time than the pattern search and that is why the pattern search is the second step in the optimization procedure, Use of the genetic algorithm is the final step in the optimization process. In Fig2, the entire three-step optimization procedure is represented as a flowchart Figue 2 Flowchart of the new grounding systems optimization seategy MI, Case Stuy A. Optimization Modeling Based on SRP Design Rules For the objective function of the grounding grid optimal design problem, there ate different functions for different purposes. According to SRP. the objective function is to ‘minimize the grounding system total construction cost. Based fon the data SRP. there are (a) the material cost of horizontal conductors and vertical rods, Cimg=S3.77/ft and CroS1.67/ft; (b) the material cast of exothermic welds, Comn-$19.25 each; (©) the labor cost of trenching. installing and backfilling conductors and driving ground rods, Cnnir$10.0/ft and Came=$32/Rt respectively; (A) the labor cost of making an exothermic connection, C.rms:-$40 each. Thus, the objective function of square and rectangular wirds is obtained using (2), 3), 4) and (5) 630 Obj. Cost N,N, Ny) = (Cos Com )* bans + (Cr+ Caee)* Log 2 *(Coona: + Cos )* Mpa * Noes) where, Laas =(W, +14, +(0, 41), ® Lowa = Iona * Nos o Nooy =(W, +1)*(N, +1) © Im above equations. L+ and L, are the given ground grid Jength and width in fl. Loon is the total length of horizontal conductors, Ln is the total length of vertical rods, and Nays is the total number of exothermic connections. In (4), hs is the length of a single vertical rod. The rod length is determined by SRP’s design rules, which are a function of the upper and lower layer soil resistivity, and upper layer depth, Rules for determining ground rod Tength are shown as follows. ‘© When the upper layer resistivity py is equal to the lower layer resistivity p:, or the depth of upper layer ht is greater than 30 f, the length of rod laa is 10 f. ‘+ When the upper layer resistivity p, is not equal to the lower layer resistivity pa, and the depth of upper layer is smaller than 10 f, the length of rod fg i820 f. ‘* When the upper layer resistivity py is not equal to the lower layer resistivity p2, and the depth of upper layer is smaller than 20 ft and greater than or equal to 10 ft. the length of rod dy is 30 f. In the optimization model, there are three additional Variables. N, is the number of meshes in grid length direction, {Ny is the number of meshes in grid width direction, and Nua is the number of rods in designed grid Based on the ground grid safety requirements and design requirements from SRP, the optimization constraints can be obtained as given by (6) (7). and (8). Ege ou S (6) foewy, o 50 te ® 508 For the square grid and rectangular grid, the number of rods is restricted as following: 4,512 © For the L shape grid, the number of rods is restricted as, following: 6

    You might also like