BCIT Citations Collection | BCIT Institutional Repository

BCIT Citations Collection

Predictive algorithm for Volt/VAR optimization of distribution networks using Neural Networks
Proceedings of IEEE Canadian Conference on Electrical and Computer Engineering (CCECE2014),May 2014, Toronto, Canada. Smart Grid functions such as Advanced Metering Infrastructure, Pervasive Control and Distribution Management Systems have brought numerous control and optimization opportunities for distribution networks through more accurate and reliable techniques. This paper presents a new predictive approach for Volt/VAr Optimization (VVO) of smart distribution systems using Neural Networks (NN) and Genetic Algorithm (GA). The proposed predictive algorithm is capable of predicting the load profile of target nodes a day ahead by employing the historical metrology data of Smart Meters, It can further perform a comprehensive VVO in order to minimize distribution network loss/operating costs and run Conservation Voltage Reduction (CVR) to conserve more energy. To test the merits of the proposed algorithm, British Columbia Institute of Technology north campus distribution grid is used as research case study., Conference paper, Published.
Retroviral genetic algorithms
Proceedings of the 2011 International Conference on Evolutionary Computation Theory and Applications. Classical understandings of biological evolution inspired creation of the entire order of Evolutionary Computation (EC) heuristic optimization techniques. In turn, the development of EC has shown how living organisms use biomolecular implementations of these techniques to solve particular problems in survival and adaptation. An example of such a natural Genetic Algorithm (GA) is the way in which a higher organism's adaptive immune system selects antibodies and competes against its complement, the development of antigen variability by pathogenic organisms. In our approach, we use operators that implement the reproduction and diversification of genetic material in a manner inspired by retroviral reproduction and a genetic-engineering technique known as DNA shuffling. We call this approach Retroviral Genetic Algorithms, or retroGA (Spirov and Holloway, 2010). Here, we extend retroGA to include: (1) the utilization of tags in strings; (2) the capability of the Reproduction-Crossover operator to read these tags and interpret them as instructions; and (3), as a consequence, to use more than one reproductive strategy. We validated the efficacy of the extended retroGA technique with benchmark tests on concatenated trap functions and compared these with Royal Road and Royal Staircase functions., Conference paper, Published.