Chapter 11 - Multifrequency electromagnetic data interpretation using particle swarm optimization and ant colony optimization methods

Abstract

Our objective is to apply and compare parameter estimation methods based on swarm intelligence and probabilistic multiagent algorithms for purposes of multifrequency electromagnetic (EM) data interpretation. To that end, particle swarm optimization (PSO) and ant colony optimization (ACO) will be used to process the multifrequency EM data. In this chapter, water saturation, water conductivity, surface conductance of clay, and contact angle of graphite are simultaneously estimated by processing multifrequency EM measurements using the PSO and ACO-based inversion methods. PSO uses a linearly decreasing inertia weight, and the ACO uses a max-min ant system. Multifrequency EM log responses from 25 distinct layers in a synthetic formation were processed by the two inversion methodologies for purposes of comparing their estimation accuracies and computational times. Both the inversion methods provide reliable estimates. ACO-based inversion takes about 50min to perform inversion for all the 25 synthetic layers, while the PSO-based inversion takes about 12min. For all the 25 layers in the synthetic formation, the average relative errors of water conductivity, surface conductance of clay, and contact angle of graphite are around 5% each and that of water saturation is around 10%. Water saturation estimates are least accurate, while contact angle estimates are the most accurate when the oil saturation is high. In order of computational speed for the inversion of multifrequency EM data, PSO is the fastest followed by simulated annealing and then ACO with Monte Carlo Markov Chain (MCMC) inversion being the slowest. PSO-based inversion is one order of magnitude faster than MCMC-based inversion. The parameters are relatively easy to tune in PSO method. We conclude that PSO inversion is the best for inverting multifrequency EM data.