Neural networks for the prediction of magnetic transformer core characteristics
TL;DR: First attempts to predict the dependence of P on several parameters of core design by means of artificial neural networks (ANN's) showed good results for simple backpropagation networks equipped with several output neurons for an adaptive version of Gaussian coarse coding.
Abstract: Because the performance of power transformers is by various distinct parameters of the magnetic core, the prediction of relevant characteristics such as no-load losses P by analytical methods is impractical. This paper reports first attempts to predict the dependence of P on several parameters of core design by means of artificial neural networks (ANN's). Investigations of several ANN versions showed good results for simple backpropagation networks equipped with several output neurons for an adaptive version of Gaussian coarse coding. A main problem arises from the fact that an increase of input parameters is linked with a large increase in training batches established by time-consuming model core experiments. As a compromise, first ANN's were trained for the prediction of the losses P/sub J/ of "linearized" joint regions as a function of the most relevant parameters, including the number of overlap steps and the mean air-gap length of joints. This yields rough estimations of the joint's contribution to the building factor for small cores. For larger cores, an ANN cascade structure was tested. It includes a second ANN that considers indirect effect of joint designs on the global distribution of losses. The major problem with an ANN-based prediction system is establishing representative training data. Modified versions of the ANN method can be applied to various tasks, including the prediction of losses and noises of full-sized cores.
Citations
More filters
TL;DR: In this article, the main technical challenges associated with the integration of wind power into power systems are discussed, including effects of wind energy on the power system, the power systems operating cost, power quality, power imbalances, power system dynamics, and impacts on transmission planning.
Abstract: Wind power is going through a very rapid development. It is among the fastest growing power sources in the world, the technology is being developed rapidly and wind power is supplying significant shares of the energy in large regions. The integration of wind power in the power system is now an issue in order to optimize the utilization of the resource and to continue the high rate of installation of wind generating capacity, which is necessary so as to achieve the goals of sustainability and security of supply. This paper presents the main technical challenges that are associated with the integration of wind power into power systems. These challenges include effects of wind power on the power system, the power system operating cost, power quality, power imbalances, power system dynamics, and impacts on transmission planning. The main conclusion is that wind power's impacts on system operating costs are small at low wind penetrations (about 5% or less). At higher wind penetrations, the impact will be higher, although current results suggest the impact remains moderate with penetrations approaching 20%. In addition, the paper presents the technology and expectations of wind forecasting as well as cases where wind power curtailment could arise. Future research directions for a better understanding of the factors influencing the increased integration of wind power into power systems are also provided.
385 citations
TL;DR: In this article, the authors conduct a literature survey and reveal general backgrounds of research and developments in the field of transformer design and optimization for the past 35 years, based on more than 420 published articles, 50 transformer books, and 65 standards.
Abstract: With the fast-paced changing technologies in the power industry, new references addressing new technologies are coming to the market. Based on this fact, there is an urgent need to keep track of international experiences and activities taking place in the field of modern transformer design. The complexity of transformer design demands reliable and rigorous solution methods. A survey of current research reveals the continued interest in application of advanced techniques for transformer design optimization. This paper conducts a literature survey and reveals general backgrounds of research and developments in the field of transformer design and optimization for the past 35 years, based on more than 420 published articles, 50 transformer books, and 65 standards.
159 citations
Additional excerpts
...The application of AI in loss evaluation is addressed in [48] and [49], where the no-load losses as a function of core design parameters are predicted by means of artificial neural networks (ANNs)....
[...]
Journal Article•
TL;DR: This paper summarizes some of the most important developments in this research area and provides a synthesis of the published research in this field to stimulate further research interests and efforts in the respective topics.
Abstract: The complexity of transformer design demands reliable and rigorous solution methods. Survey of current research reveals the continued interest in application of advanced techniques for transformer design optimization. This paper summarizes some of the most important developments in this research area. The main purpose is to provide a synthesis of the published research in this field and stimulate further research interests and efforts in the respective topics.
51 citations
05 Jun 2006
TL;DR: A hybrid artificial intelligence/numerical technique is proposed for the selection of winding material in power transformers that uses decision trees and artificial neural networks for winding material classification, along with finite-element/boundary element modeling of the transformer for the calculation of the performance characteristics of each considered design.
Abstract: The aim of the transformer design optimization is to define the dimensions of all the parts of the transformer, based on the given specification, using available materials economically in order to achieve lower cost, lower weight, reduced size, and better operating performance. In this paper, a hybrid artificial intelligence/numerical technique is proposed for the selection of winding material in power transformers. The technique uses decision trees and artificial neural networks for winding material classification, along with finite-element/boundary element modeling of the transformer for the calculation of the performance characteristics of each considered design. The efficiency and accuracy provided by the hybrid numerical model render it particularly suitable for use with optimization algorithms. The accuracy of this method is 96% (classification success rate for the winding material on an unknown test set), which makes it very efficient for industrial use
43 citations
TL;DR: Experimental results show that NSGA-II+FEM model successfully provides a global feasible solution by minimizing total loss and related cost while improving the efficiency of three-phase transformer, rendering it suitable for application in the design environment of industrial transformers.
Abstract: In order to obtain a good optimization method for the electrical transformer design with optimal selection of parameters, performance evaluation of three evolutionary algorithms (EAs), namely, genetic algorithm (GA), differential evolution algorithm, and nondominated sorting GA (NSGA-II), is carried out. The aim of this paper is to optimize parameters of transformer design (core thicknesses, primary-turn number, secondary-turn number, primary conductor area, and secondary conductor area) for minimization of total power losses (no-load losses and load losses) in three-phase transformer topology while maintaining high efficiency and low cost. The method used for this optimization scheme combines the finite-element method (FEM) and EAs to provide an accurate selection of parameters together with the optimized magnetic flux density and decreased loss. Experimental results show that NSGA-II+FEM model successfully provides a global feasible solution by minimizing total loss and related cost while improving the efficiency of three-phase transformer, rendering it suitable for application in the design environment of industrial transformers.
36 citations
Cites background from "Neural networks for the prediction ..."
...means of artificial neural network (ANN) as in [8] and [9]....
[...]
References
More filters
13 Apr 1992
TL;DR: The procedure explained can be used to provide good initial designs for use with iterative search techniques to reduce searching time and is highly desirable in connection with increasing the effectiveness of the optimal design procedure.
Abstract: A novel method for the optimal design of the electromagnetic devices is presented The method utilizes artificial neural networks (ANNs) in a design environment which encompasses numerical computations and expert's input for generating a variety of ANN training data Results of two implementation examples are provided The optimal design is obtained quickly (in a matter of milliseconds) once the ANNs are trained with a variety of geometrical topologies The procedure explained can be used to provide good initial designs for use with iterative search techniques (currently used) to reduce searching time This aspect is highly desirable in connection with increasing the effectiveness of the optimal design procedure >
59 citations
"Neural networks for the prediction ..." refers methods in this paper
...A first attempt for the establishment of an ANN in the field of magnetic circuits was to optimize a simple pot core [ 9 ]....
[...]
TL;DR: In this article, the authors investigated the ac-permeability of transformer core components and found that the low ratio /spl mu//sub z/ yields a tendency of constant package flux throughout the whole core.
Abstract: In spite of extensive optimizations of transformer core designs, investigations of full sized cores showed distinct inhomogeneities of flux density B. Limbs showed discontinuous variations of B in peripheral packages and minima of B in thick central ones. The latter are not caused by global eddy currents but rather by localized flux components /spl Phi//sub z/ normal to the sheet plane. Attempts to determine the respective effective ac-permeability /spl mu//sub z/ yielded values below 100, i.e., almost three orders below /spl mu//sub x/ of the rolling direction. For given B, (planar) eddy current losses P/sub z/ proved to exceed the respective values P/sub x/ by two orders, a ratio which increases with increasing length L of the magnetized sheet region. The low ratio /spl mu//sub z///spl mu//sub x/ yields a tendency of constant package flux throughout the whole core. A key criterium for /spl Phi//sub z/-components between packages proved to be the overlap regions which were studied in a comparative way for several step-lap configurations. Distinct differences of respective values of lap-region excitation V/sub L/ were observed as a function of air gap lengths and the step number N, respectively. Variations of V/sub l/-and especially overlaps of high V/sub L/ in connection with shifted overlap regions-proved to yield /spl Phi//sub z/-components including flux transfer between packages. In a complex way, shifts yielded decreasing excitation power, but increased core losses due to planar eddy currents. In addition, package shifts cause both the discontinuities of B of thin peripheral packages and the minima of local B in thick central ones. With respect to core design, it can be assumed that small shifts favor take over of flux without causing significant planar eddy current losses (due to small L), while large shifts increase total losses in a disadvantageous way. With increasing N, these effects become less significant. >
46 citations
"Neural networks for the prediction ..." refers background or result in this paper
...This quantity has proved to be very significant in an earlier study [ 6 ]....
[...]
...On the other hand, low yields a flux concentration along the inner edge of the core limbs and yokes and thus inhomogeneous flux density throughout the core [ 6 ]....
[...]
...In addition, for effortless establishment of data batches for first training attempts, and were determined at overlap regions assembled in a “linearized way” [ 6 ], [13], in most cases....
[...]
...Based on a concept previously discussed in [ 6 ], the losses and the excitation of the core’s “bulk” parts are simply calculated by a conventional estimation unit (CEU) assuming nominal values for limbs and yokes (see Section V-A)....
[...]
...On the other hand, the overlap regions of joints have additional losses which are represented by the “joint loss” . As suggested earlier [ 6 ], we express the latter by area related joint loss (in W/m2) for a core with four corners where an overlap design exists....
[...]
TL;DR: The paper describes the present status of the numerical simulation of flux and loss distributions in electrical machinery, mainly transformers.
Abstract: In recent years, techniques of numerical analysis have rapidly progressed. By introducing these techniques, the design of smaller and more efficient electrical machinery becomes easier without repeating trial manufacture which is time consuming, laborious and money wasteful. The paper describes the present status of the numerical simulation of flux and loss distributions in electrical machinery, mainly transformers.
30 citations
"Neural networks for the prediction ..." refers methods in this paper
...easily modeled by finite-element analysis [ 13 ] (for drawbacks see Section V)....
[...]
...In addition, for effortless establishment of data batches for first training attempts, and were determined at overlap regions assembled in a “linearized way” [6], [ 13 ], in most cases....
[...]
01 Nov 1995
TL;DR: In this paper, it is shown that an increase of the number of N of steps reduces the sensitivity of the core in respect to variations of g and s. This is due to increasing critical induction BC as of which the excitation VL of the overlap region steeply increases.
Abstract: It has been stated that multistep-lap (MSL) jointed cores show distinct improvements regarding power losses P in comparison to single-step-lap (SSL). On the other hand, model core experiments tend to yield contradictory results. In the paper, special emphasis was put on the underlying physical mechanisms, taking into consideration that model cores usually show 'perfect' stacking, while full-size cores exhibit considerably high air gap lengths g as well as shifts s between the overlap regions of adjoining packages. It is shown that an increase of the number N of steps reduces the sensitivity of the core in respect to variations of g and s. This is due to increasing 'critical induction' BC as of which the excitation VL of the overlap region steeply increases. Low VL corresponds to lower balancing interlaminar flux in the overlap/gap regions and thus to a generally more homogeneous flux distribution linked with lower P. The quantity BC can be assumed to be of high practical relevance for the design of transformer cores.
28 citations
"Neural networks for the prediction ..." refers background in this paper
...c) For the narrow core, the joints’ contribution to excitation is small as long as is below its critical value which according to [ 7 ] is 1 T for (G4, G8), 1.5 T for (G2), and 1.6 T for (G6)....
[...]
...For an ANN of Type III, Fig. 13 shows examples of the predicted improvement which can be attained by replacing SSL ( ) by MSL ( ). The predictions reflect the industrial finding that improvements can be expected only for induction values which exceed the so-called critical induction , corresponding to saturated air gap bridges, which is close to 1 T for [ 7 ]....
[...]
...Apart from decreasing audible noise [3], [4], MSL tends to yield a reduction of losses by several percent [5]‐[ 7 ]....
[...]
TL;DR: In this paper, the authors discuss the respective relevance of overlap length, air gap length, and length of shifts of overlap regions of adjacent core packages for single and multi-step lap.
24 citations
"Neural networks for the prediction ..." refers background in this paper
...Apart from decreasing audible noise [3], [ 4 ], MSL tends to yield a reduction of losses by several percent [5]‐[7]....
[...]