Investigation on the Spatial Variation of 1-D Velocity Models in a Geothermal Environment using a Genetic Algorithm

The analysis of microseismicity in geothermal fields is used as a tool to provide information for reservoir characterisation and to monitor the response to development operations. A particular challenge in the determination of earthquake hypocentre locations is the high degree of heterogeneity of reservoir conditions (e.g. rock formation, pressure, temperature, fluid phase etc.) resulting in complex seismic velocity variations. For manageable computation, these variations generally need to be represented by a 1-D velocity model. An accurate 1-D model is critical not only for earthquake hypocentre locations but also meaningful inversions of 3D velocity structure using tomographic methods. In this study, an improved characterisation of seismic velocity variations is attempted both for processing and interpretation of microearthquake data with application to the Wairakei geothermal field, New Zealand. A genetic algorithm (GA) can be used to derive a minimum 1-D velocity model if little about the velocity structure is known. This is often the case in a geothermal field where reservoir information is scattered and direct measurements from active seismic exploration surveys are rare, ineffective or unavailable. GAs are effective in how the data parameter space is explored, because a group of velocity models that are randomly chosen are handled simultaneously and tested for their data fitness. GAs are superior to other random search techniques in that during a stochastic process based on Darwin’s laws of natural evolution, the parameter search is guided towards a globally optimal solution. We derive several 1-D velocity models for the Wairakei geothermal field by subdividing the earthquake data in the area of interest to explore velocity variation throughout the field. In addition, we studied the range of obtained models (an ensemble) for each specific area, to better understand the degree of similarity of these models or if the properties of certain zones remain unconstrained. This aims at a better understanding of non-uniform uncertainty on the obtained structures and trade-offs between velocities and layer thicknesses. Our study shows that the GA velocity inversion can capture and quantify spatial velocity variation in a geothermal environment.