Minimization of voltage deviation and power losses in power networks using Pareto optimization methods

https://doi.org/10.1016/j.engappai.2010.01.011Get rights and content

Abstract

Voltage regulation is an important task in electrical engineering for controlling node voltages in a power network. A widely used solution for the problem of voltage regulation is based on adjusting the taps in under load tap changers (ULTCs) power transformers and, in some cases, turning on Flexible Alternating Current Transmission Systems (FACTS), synchronous machines or capacitor banks in the substations. Most papers found in the literature dealing with this problem aim to avoid voltage drops in radial networks, but few of them consider power losses or meshed networks. The aim of this paper is to present and evaluate the performance of several multi-objective algorithms, including hybrid approaches, in order to minimize both voltage deviation and power losses by operating ULTCs located in high voltage substations. In particular, a well-known multi-objective algorithm, PAES, is used for this purpose. PAES finds a set of solutions according to Pareto-optimization concepts. Furthermore, this algorithm is hybridized with simulated annealing and tabu search to improve the quality of the solutions. The implemented algorithms are evaluated using two test networks, and the numerical results are analyzed with two metrics often used in the multi-objective field. The results obtained demonstrate the good performance of these algorithms.

Introduction

The operation of power networks has undergone considerable change in recent times. Several factors have influenced this change. For example, the interest in saving energy, the use of cheaper and more universal measuring devices, the growing demand from consumers for improved quality of the power supply, or even the appearance of new tools for monitoring power networks. New technologies are constantly appearing in all fields of science and engineering, including electrical engineering and power systems. Specifically, there are many elements in power distribution networks that could be automated and coordinated, such as tap changers in power transformers or tap changers in bank capacitors. The present paper deals with new computational optimization algorithms which allow a better operation of the system, achieving a better voltage profile and reduced power losses. The general problem of voltage control and optimization of power flow occurs because of the need to act on the voltage at the network nodes to ensure that it remains within acceptable limits. One of the main elements allowing this adjustment is the under load tap changer (ULTC), which can adjust the voltage ratio in discrete steps. Taking into account the combinatorial nature of this problem, different methodologies based on artificial intelligence have been applied. Some authors have proposed the use of techniques such as dynamic programming (Lu and Hsu, 1995) to control tap changers and capacitor banks in substations for 24-hour-planning. This aims to improve the efficiency of the algorithm by limiting the search space in the vicinity of an analytical solution calculated using the hourly load forecasting. This technique has even been extended to small networks with radial capacitor banks at the heads of the lines (Ruey-Hsun and Chen-Kuo, 2001). Fuzzy logic techniques have also been used (Liu et al., 2002). Other authors have used neural network techniques for voltage and power regulation in radially operated networks during one-day periods, making use of the tap changers of transformers and capacitor banks (Saric et al., 1997). Other techniques such as genetic algorithms (Haida and Akimoto, 1991) have also been implemented for voltage optimization. On the other hand, the typical problem of voltage and reactive power compensation (known as the Volt/Var problem) has been widely studied in power networks (Begovic et al., 2004).

Moreover, the variation of voltage at the system nodes due to ULTCs could involve a change in the power flow through the system lines, which must be minimized in order to avoid the power losses that occur due to the Joule effect. In this way, optimal operation of ULTCs allows voltage and power flow to be under control. Thus, this problem can be treated from a multi-objective perspective (Augugliaro et al., 2004). The traditional way of dealing with multi-objective optimization problems consists of introducing weights in a simple aggregating function (weighted sum), where weights reflect the relative significance of the different objectives. However, the main drawback of this linear combination is the difficulty to establish accurate values for the weights, especially when the objectives to be considered have different scales or represent different magnitudes. An interesting, and possibly the best way of overcoming this drawback is the use of Pareto optimization (Deb, 2001), which is a method that includes multiple criteria without using weights. This paper addresses the problem of optimizing the control of an automated power distribution network under normal operation using Pareto-optimization, where the objectives to minimize are: a voltage deviation index that gives information about the voltage profile in the network (Montoya et al., 2006), and the power loss of electrical lines due to the Joule effect.

Section snippets

Voltage regulation in electrical power networks

Currently, voltage regulation in traditional power substations for HV/MV (high/medium voltage) or MV/MV transformers is performed by under load tap changers (ULTC), whose input variables are essentially the voltage and current values measured at the outlet of the secondary circuit of the transformer, while regulation in the MV/LV (medium/low voltage) is carried out manually using vacuum tap changers. In an automated network, this process can be approached from a different, more holistic

Results and discussion

When applying heuristic methods to optimization problems it is advisable to perform a sensitivity analysis, i.e. to determine to what extent the output of the model depends upon the inputs. These sensitivity analyses require a certain number of executions of each algorithm in order to obtain robust conclusions. Thus, it is important to note that while in single-objective optimization it is straightforward to determine the median value of several runs, in Pareto-based multi-objective

Conclusions

This paper addresses the optimization problem consisting of adjusting ULTC transformers in order to regulate nodal voltages in power distribution networks and simultaneously minimize power losses. The problem is of great interest for power distribution companies because this strategy allows them to act on the main elements of the network to determine the optimal voltage profile with the aim of correcting possible deviations from nominal values and to reduce inefficiencies. At the same time

Acknowledgments

This work was partially supported by grants TIN2008-01117 (Spanish Ministry of Science and Innovation), and P2007-TIC-02988 (CICE, Junta de Andalucía).

References (20)

  • Y. Liu et al.

    Optimal volt/var control in distribution systems

    International Journal of Electrical Power and Energy Systems

    (2002)
  • A. Saric et al.

    Neural-net based coordinated control of capacitors and ULTC transformer in daily operation of radial distribution systems

    Electric Power Systems Research

    (1997)
  • A. Augugliaro et al.

    Voltage regulation and power losses minimization in automated distribution networks by an evolutionary multiobjective approach

    IEEE Transactions on Power Systems

    (2004)
  • R. Baños et al.

    A hybrid meta-heuristic for multi-objective optimization: MOSATS

    Journal of Mathematical Modelling and Algorithms

    (2007)
  • Begovic, M.M., Radibratovic, B., Lambert, F.C., 2004. On multiobjective volt-VAR optimization in power systems. In:...
  • F. Glover et al.

    Tabu Search

    (1997)
  • D.E. Goldberg

    Genetic algorithms in search, optimization, and machine learning

    (1989)
  • A. Gómez-Expósito et al.

    Electric Energy Systems: Analysis and Operation

    (2008)
  • C. Grigg et al.

    The IEEE reliability test system-1996. A report prepared by the reliability test system task force of the application of probability methods subcommittee

    IEEE Transactions on Power Systems

    (1999)
  • Haida, T., Akimoto, Y., 1991. Genetic algorithms approach to voltage optimization. In: Proceedings of the First...
There are more references available in the full text version of this article.

Cited by (48)

  • Complex network community detection method by improved density peaks model

    2019, Physica A: Statistical Mechanics and its Applications
    Citation Excerpt :

    Generally, complex network is simplified into several communities, and nodes with common or similar features are assigned to the same community. In the real world, a large number of practical cases can be considered as complex network structures, such as mobile communication networks, bioinformatics [1], power system networks [2,3], Internet, social network [4], information systems [5,6], physics [7]. If complex network structure is clear and concise, the performance analysis of network can achieve optimal results or balance status quickly.

  • Adaptive community detection in complex networks using genetic algorithms

    2017, Neurocomputing
    Citation Excerpt :

    Graphs are used to model real systems in many areas, such as power systems [16], sociology [6], biology [23], physics [1], and informatics [2,8] thanks to the computers, which allow to efficiently representing, managing and processing large amounts of data, including graph-based structures. Among other emerging scientific areas, the analysis of communities in complex systems has gained importance due to their application to multiple contexts, including collaboration networks, biological systems, social networks in the Internet, transportation networks, or electrical networks [12]. All complex systems share a common characteristic: community structures [14] that consist of groups of nodes inside a network that are more densely connected than with the remaining nodes of the network.

  • Higher-order statistics for power systems: Effects of the sampling frequency on ergodicity

    2016, Applied Mathematical Modelling
    Citation Excerpt :

    Other methods, models, or algorithms should also be mentioned, including MUSIC, ESPRIT, and Kalman filtering [16]. Furthermore, techniques from the field of multi-objective optimization have proved useful for addressing power quality problems [24]. In this study, we propose the use of HOS for harmonic analysis based on properties P1–P3.

View all citing articles on Scopus
View full text