Estimating technical losses is fundamental to the planning and economics of electric power networks. This paper surveys the evolution of the ideas behind energy loss estimation and focuses on the development of the concepts of the loss factor and equivalent hours. The paper next identifies difficulties in using maximum demands and the loss factor to estimate energy losses. Based on this analysis, this study proposes an alternative loss estimation approach that relies on the “loss coefficient” as the fundamental parameter for describing load variations in loss estimation. A large load-curve data bank from Brazilian utilities is used to characterize load-curve parameters and provide perspective on the old and new concepts. Practical applications put the proposed ideas into perspective, showing how the use of average demands and loss coefficient can help to make better cable choices, increase accuracy in loss estimation for distribution transformers, and enhance the quality of information in loss estimation analysis.

Reliability analysis of power systems has been attracting increasing attention. Regulatory agencies establish reliability standards that, if infringed, result in costly fines for the utility suppliers. A special concern pertains to the distribution networks on which most failures occur. The allocation of switches is a possible strategy to improve reliability, by allowing network reconfiguration to isolate contingencies and restore power to dark areas. This paper proposes an optimization methodology to allocate switches on radially operated distribution networks. The solution framework considers sectionalizing and tie switches of different capacities, with manual or automatic operation schemes. The approach minimizes the costs of allocation and energy not supplied, under reliability and flow capacity constraints. The solution framework is based on memetic algorithm concepts with a structured population. Case studies with a large network and real-world scenarios were used to evaluate the methodology. The results indicate that significant cost reductions can be achieved using the proposed solutions.

In power distribution networks the load varies within any given time frame. It may, therefore, seem that a good approach to reduce losses would be the solving of a network reconfiguration problem to suit each of the significant load variations. However, frequent changes in configuration can trigger outages or cause transient problems; they are best avoided. A recent formulation of this problem explicitly considers load variations and proposes to restrain frequent reconfigurations by assuming that network topologies will remain unchanged for a given planning period. This formulation leads to a much larger optimization problem than that traditionally used for network reconfiguration; moreover, it requires a new approach to optimization which is capable of dealing with energy flows instead of only instantaneous power flows. Such an approach is proposed in this paper, which discusses the design of an adaptive hybrid genetic algorithm that fulfills these new requirements. Key concepts in evolutionary computation and analysis of distribution systems are explored to develop this new algorithm. Application to real case studies certifies its benefits.

In this paper decomposition and coordination techniques are applied to minimize the operation al cost of hydrothermal power systems. Through this for muXation large systems are dealt with by a more general and precise model that includes thermal operation costs and hydroelectric power generation as nonlinear functions, water head variations, stochastic load demands, hydraulic networks with cascade plants, time delays, and spilling. An economic interpretation and an example are given.

Works requiring taxonomic knowledge face several challenges, such as arduous identification of many taxa and an insufficient number of taxonomists to identify a great deal of collected organisms. Machine learning tools, particularly convolutional neural networks (CNNs), are then welcome to automatically generate high-performance classifiers from available data. Supported by the image datasets available at the largest online database on ant biology, the AntWeb (www.antweb.org), we propose here an ensemble of CNNs to identify ant genera directly from the head, profile and dorsal perspectives of ant images. Transfer learning is also considered to improve the individual performance of the CNN classifiers. The performance achieved by the classifiers is diverse enough to promote a reduction in the overall classification error when they are combined in an ensemble, achieving an accuracy rate of over 80% on top-1 classification and an accuracy of over 90% on top-3 classification.