Darko Capko

Algorithms in electric power system one-line diagram creation

This paper analyzes various algorithms for the automatic generation of electric power system one-line diagrams. Historically these one-line diagrams were created manually. As these diagrams can be large, their manual creation takes a lot of time and resources, and it is also prone to errors. One possible solution to this problem is the automatic generation of one-line diagrams. Automatic visualization can be performed by existing rule-based algorithms, or if the network is modeled by a mathematical graph, then the problem can also be solved by graph drawing algorithms. Initial results achieved with soft computing algorithms will also be shown. We discuss the similarities, differences and applicability of these different algorithms in automatic one-line diagram generation. The final goal is to find the most suitable algorithm for the automatic generation of visually pleasing one-line diagrams, which allow dispatchers and engineers working in control centers a higher level of efficiency in performing their everyday tasks.

Distributed PSO Algorithm for Data Model Partitioning in Power Distribution Systems

This paper presents a method for data model partitioning of power distribution network. Modern DistributionManagement Systems which utilize multiprocessor systems for efficient processing of large data model areconsidered. The data model partitioning is carried out for parallelization of analytical power calculations. The proposedalgorithms (Particle Swarm Optimization (PSO) and distributed PSO algorithms) are applied on data model describinglarge power distribution network. The experimental results of PSO and distributed PSO algorithms are presented.Distributed PSO algorithm achieves significantly better results than the basic PSO algorithm.

A HYBRID GENETIC ALGORITHM FOR PARTITIONING OF DATA MODEL IN DISTRIBUTION MANAGEMENT SYSTEMS

In this paper, we propose a Hybrid Genetic Algorithm for data model partitioning of power distributionnetwork. Analytical functions are the core of Distribution Management Systems (DMSs). Efficient calculation of thefunctions is of the utmost importance for the DMS users; the necessary preconditions for the efficient calculation areoptimal load balancing of processors and data model partitioning among processors. The proposed algorithm is appliedto different real models of power distribution systems. It obtains better results than classical evolutionary algorithms(Genetic Algorithm and Particle Swarm Optimization). The Hybrid Genetic Algorithm also achieves better results thanmultilevel algorithm (METIS) in cases of small graphs. http://dx.doi.org/10.5755/j01.itc.40.4.981

An optimal relationship-based partitioning of large datasets

Modern adaptive applications utilize multiprocessor systems for efficient processing of large datasets where initial and dynamic partitioning of large datasets is necessary to obtain an optimal load balancing among processors. We applied evolutionary algorithms (Genetic Algorithm and Particle Swarm Optimization) for initial partitioning, and diffusion (DR) and cut-and-paste (CP) algorithms for dynamic partitioning. Modified versions of DR and CP algorithms are developed to improve dynamic partitioning running in NUMA multiprocessor systems. The proposed algorithms were applied on datasets describing large electricity power distribution systems and experimental results prove reductions of processor load imbalance and performance improvements.

Architecture of Real-Time Database in Cloud Environment for Distributed Systems

In every distributed system there is a need for a real-time database when working with a large amount of data and when quick response from a service is expected on the client side. Cloud companies do not offer real-time databases, therefore, in this paper we present the architecture of a real time database in cloud environment, as well as reviews of other papers dealing with this topic, analyzing their advantages and disadvantages. The above-mentioned real-time database attempts to store different kind of data, and our proposed solution is implemented in each distributed system service. Unlike other solutions, this real-time database will not occupy main memory-base because it is running in cloud environment. Also, we are proposing which cloud environment services can be used to implement this real time database. As an example, we are modeling one distributed system using Petri nets where solution can be applied.

Electric power distribution system visualization with graph partitioning

This paper presents an electric power distribution system (EPDS) visualization algorithm which performs graph partitioning before generating the visual representations, i.e. one line diagrams. The proposed algorithm is different from similar algorithms as it performs graph partitioning to reduce the solution space. The partitioning step is necessary because modern EPDS can consist of thousands of distribution transformers, pole switches and line segments. After the partitioning step, in its second phase, the algorithm lays out the subgraphs and then combines the drawings into complete one-line diagrams. During the experiments, the partitioning and layout phases were performed with various combinations of the genetic and branch and bound algorithm. The algorithm was tested on partial, real life electric power distribution systems of up to 150 objects and it generated visually pleasing one-line diagrams.

Adaptive neural network workflow management for Utility Management Systems

This paper focuses on grid performance optimization in large scale workflow applications with an intelligent workflow scheduling mechanism. Utility Management Systems (UMS) are managing very large numbers of workflows with very high resource requirements. This paper proposes a UMS scheduling architecture which dynamically executes a scheduling algorithm using near real-time feedback about the current status of grid nodes. Workflow scheduling was performed with an artificial neural network (ANN). The network was trained in a system with three workflows. The case study presented in this paper shows results achieved in a three workflow system, as well as results achieved in a five workflow system where an adaptive ANN was used. The results testify that significant improvement of overall execution time can be achieved by adapting weights in the neural network.