Adaptive Hybrid Genetic Algorithm for Technical Loss Reduction in Distribution Networks Under Variable Demands
Abstract
In power distribution networks the load varies withinany given time frame. It may, therefore, seem that a good approachto reduce losses would be the solving of a network reconfigurationproblem to suit each of the significant load variations. However,frequent changes in configuration can trigger outages or causetransient problems; they are best avoided. A recent formulationof this problem explicitly considers load variations and proposesto restrain frequent reconfigurations by assuming that networktopologies will remain unchanged for a given planning period.This formulation leads to a much larger optimization problemthan that traditionally used for network reconfiguration; moreover,it requires a new approach to optimization which is capableof dealing with energy flows instead of only instantaneous powerflows. Such an approach is proposed in this paper, which discussesthe design of an adaptive hybrid genetic algorithm that fulfillsthese new requirements. Key concepts in evolutionary computationand analysis of distribution systems are explored to developthis new algorithm. Application to real case studies certifies itsbenefits.
Index Terms—Distribution of electric power, genetic algorithms,hybrid genetic algorithms, loss reduction, network reconfiguration,technical losses, variable demands.
I. INTRODUCTION
AREDUCTION of technical losses in electric power distributionnetworks can be regarded as a source of energy.Power engineers constantly strive to harness this energysource by using the best possible strategies in network planningand operation, such as network reconfiguration and capacitorallocation.This paper focuses on the network reconfiguration problemof finding a topology for the primary distribution network thatwill minimize technical losses throughout a given planning period.Special attention is given to the variation in loads duringthe entire planning period. Its main contribution is the designof an adaptive hybrid genetic algorithm capable of dealing withthe representation of the problem proposed by Bueno, Lyra and Cavellucci [1]. This representation formally considers demandvariations and provides a solution for the dichotomy of “good”and “bad” inherent in the problem: good loss reduction, bad frequentreconfigurations. On the downside, this new representationsubstantially increases the size of the associated optimizationproblem. Moreover, it requires the design of optimizationapproaches that consider the total energy flow throughout theplanning period.For economic reasons, most power distribution systems operatewith a radial configuration. Normally closed switches arelocated on the live path from substations to load points, whereasnormally open switches are located off the live path and are designedto provide alternative network configurations for copingwith emergencies, maintenance procedures and changes in loadpatterns.More than 30 years ago, the French engineers A. Merlin andH. Back [2] perceived an opportunity to reduce technical lossesby exploring a change in the status of normally closed and normallyopen switches. They proposed the “network reconfigurationproblem”, for which the solution should provide the “best”status for all the switches in a primary distribution network, bestin the sense that they provide a radial configuration supplyingloads with the minimum of power loss. However, because thenumber of possible solutions for the network reconfigurationproblem grows exponentially with the number of switches, itis difficult to find an optimal solution when the size of networkincreases. Merlin and Back relied on formal analysis and engineeringinsights to devise an approximate method, which providedgood solutions for the problem but with no guarantee ofglobal optimality.A vast bibliography on the network reconfiguration problemis now available. It includes attempts to find a global optimalsolution [3], new approximate methods [4], [5] and alternativeparadigms of approaching the problem, such as evolutionarycomputation [6] and other bio-inspired techniques [7]. Mostof these approaches consider fixed demands at each load point(usually the maximum demands forecast for the planning period).However, loads vary during any given planning period,with a different pattern for each bus, and these variations mustbe considered in the search for a minimum loss configuration.


Download full report
http://ieeexplore.ieeeiel5/59/4762160/04...er=4762173