Pedro Pereira

This work deals with the multi-objective optimization of analog circuits by generating the Pareto front where elements are low sensitive to parameters' variations. NSGA-II is used for obtaining the non-dominated solutions. Richardson extrapolation technique is used for the in-loop optimization approach for computing partial derivatives and, thus, the solutions' sensitivity. NSGA-II Pareto fronts' intrinsic ranking is exploited for the generation of the new ‘low-sensitive’ Pareto front. The case of the optimal sizing of a CMOS voltage follower is considered to exemplify the proposed approach.

The goal of this paper is to present a comparison among three known metaheuristics: Genetic Algorithm (GA), Particle Swarm Optimization (PSO) and Simulated Annealing (SA). For the comparison, the design of an LC - Voltage Controlled Oscillator (LC-VCO) is considered, where the minimization of both VCO phase noise and power consumption is envisaged. The objective of this comparison is to find the algorithm yielding the best solution. The validity of the solution obtained with each metaheuristic algorithm is checked against HSPICE/RF simulation results. Robustness checks for each algorithm are presented at the end of this paper.

Inductive superconducting fault current limiters have already demonstrated their technical viability in electrical networks. Its architecture and robustness make them potentially adequate for distribution networks, and this type of devices is considered as an enabling technology for the advent of embedded generation with renewable energy sources. In order to promote the growth and maturity of these superconducting technologies, fast design tools must be developed, allowing simulating devices with different materials in grids with diverse characteristics. This work presents advances in the development of such tool, which, at present stage, is an effective alternative to software simulations by finite elements methods, reducing dramatically computation time. The algorithms are now compared with experimental results from a laboratory scale prototype, showing the need to refine them.