Leonidas C. Resende

Composite Systems Reliability Evaluation Based on Monte Carlo Simulation and Cross-Entropy Methods

This paper proposes a new approach to evaluate loss of load indices in composite generation and transmission systems. The main idea is to combine a Cross-Entropy (CE)-based optimization process and nonsequential Monte Carlo Simulation (MCS) to obtain an auxiliary sampling distribution, which can minimize the variance of the reliability index estimators. This auxiliary sampling distribution will properly modify the original unavailabilities of both generation and transmission equipment, so that important failure events are sampled more often. As a result, the MCS algorithm can reach convergence faster and with fewer samples, leading to significant gains in computational performance, especially when dealing with very reliable system configurations. The proposed method is tested using several composite power systems, including the IEEE RTS 79, IEEE RTS 96, and a configuration of the Brazilian system.

Evolution Strategies to Transmission Expansion Planning Considering Unreliability Costs

This paper presents a new methodology to solve transmission expansion planning (TEP) problems based on evolution strategies (ES), but other heuristics are also used to assist the search process. The TEP problem includes the search for the least cost solution, bearing in mind investments and interruption costs. Unreliability costs are considered through the index LOLC-loss of load cost. Moreover, the dynamic nature of the TEP is accounted for by the proposed methodology. Case studies on a small test and on a real sub-transmission network (CEMIG Company, Brazil) are presented and discussed

Application of Monte Carlo Simulation to Well-Being Analysis of Large Composite Power Systems

This paper presents a new methodology to evaluating the well-being indices of large composite generation and transmission systems. A well-being framework is used to classify the system states into healthy, marginal and at risk, according to a pre-defined deterministic criterion. In order to combine deterministic and probabilistic concepts, the proposed methodology uses a non-sequential Monte Carlo simulation, a multi-level non-aggregate Markov load model and test functions to estimate the well-being indices for the system and load buses. Moreover, a network reduction is also proposed to find an equivalent well-being framework suitable to practical large power systems. Case studies on an IEEE standard system and on a configuration of the Brazilian network are presented and discussed

Composite Reliability Assessment of Power Systems with Large Penetration of Renewable Sources

The constant increase in oil prices and the concern over the reduction of gas emissions causing the greenhouse effect favor the creation of policies to encourage the production of energy through renewable sources. The recent restructuring of the electricity sector has introduced new concepts such as power market, transmission open access, cogeneration, independent production, etc., which enabled the decentralized energy generation, strengthening such policies. Thus, non-conventional energy sources, namely wind power, mini-hydro, solar, and cogeneration (e.g., biomass), start having a significant contribution in the energy production matrix. However, if the volatility of the available capacity from such sources is not properly considered, the decisions taken in power systems expansion and/or operation planning can severely endanger the reliability of the power supply. Thus, systems planners and operators will require new computational tools capable of coping with these characteristics, in addition to the recent power system market implementation in a deregulated environment.

Support Vector Machine application in composite reliability assessment

This paper presents a methodology for assessing the reliability indices for composite generation and transmission systems based on Support Vector Machines (SVM). The importance of SVMs is its high generalization ability. The SVMs are used to classify data into two distinct classes. These can be named positive and negative. Thus, the basic idea is to classify the system states into success or failure. For this, a pre-classification of states is achieved by performing the proposed SVM-based neural network, where the sampled states during the beginning of the non-sequential Monte Carlo simulation (MCS) are considered as input data for training and validation sets. By adopting this procedure, a large number of states are classified by a simple evaluation of the network, providing significant reductions in computational costs. The proposed methodology is applied to the IEEE Reliability Test System and to the IEEE Modified Reliability Test System.

Constructive heuristic algorithm for sub-transmission system planning

In this paper, a new methodology to assist system planners is proposed, which is able of providing a set of good solutions to the addition of new transmission lines, in order to reinforce the sub-transmission network. To circumvent the combinatorial explosion of reinforcement alternatives, a constructive heuristic algorithm is used for adding new branches to the sub-transmission grid. Based on performance indices to measure the attractiveness of new branches and using an expansion tree, the proposed algorithm is able to capture the combined effect of reinforcement additions. For selecting the best alternative, from the set of good reinforcement solutions, the planned network performance in meeting the future load demand is evaluated, considering the following operational aspects: ohmic losses, reliability, branch loading condition, and voltage profile. The evaluation of the performance indices and all operational aspects is achieved by applying a non-linear AC model-based power flow. For the reliability assessment, an enumeration technique is used to select system states. A large sub-transmission Brazilian system belonging to CEMIG utility is used to validate the methodology and the results are presented and extensively discussed.