Research And Application Of Micro-seismicity-constraint Tracer Test Data Inversion Methodology System For Hydraulically Enhanced Fracture Network Characterization In The Hot Dry Rock

The hydraulically stimulated fracture network in the Hot Dry Rock（HDR）directly affects the flow and heat transfer process of circulating fluid,thus controlls the thermal recovery efficiency,operation life and economic benefits of the EGS project.The fracture network is too complex to be characterized using a single source of monitoring data,such as well logging,induced micro-seismicity or tracer test.It is taken as the subject in this research.Based on the technical principles of induced micro-seismicity and tracer test,and the response characteristics of these monitoring data,the micro-seismicity-constraint tracer test data inversion methodology system for hydraulically stimulated fracture network characterization in the HDR is proposed,to deterministically and efficiently inverse the fracture network parameters.The main research contents and conclusions are as follows:（1）The spatial and temporal distribution of induced micro-seismicities is indicative of the disturbance characteristics of pore pressure during the hydraulic-stimulation,based on which the hydraulic diffusivity coefficient can be calculated,providing deterministic constraints for the equivalent permeability and the fracture network parameters,including fracture intensity,equivalent radius and hydraulic apertures.Therefore,based on the correlation between the monitoring data of spatial and temporal characteristics of micro-seismicity and tracer test and fracture network parameters,the micro-seismicity-constraint tracer test data inversion methodology system for fracture network characterization in the HDR is proposed and devoloped,including:the micro-seismicity-constraint tracer test data inversion method based on a heterogeneous equivalent porous medium（EPM）model for analysis of spatial distribution of permeability and the micro-seismicity-constraint tracer test data inversion method based on a discrete fracture network（DFN）model for more precise characterization of the fracture network.The Long Short-Term Memory（LSTM）neural network model is applied as the surrogate for the forward numerical models of hydraulic stimulation and tracer test.Its single inversion iteration takes less than 0.2 s.The Harris inversion algorithm is integrated with the LSTM neural network model to efficiently inverse fracture network parameters.（2）The Habanero EGS in Australia with detailed and published monitoring data was selected for application and verification of the above fracture network characterization methodology system.Based on the site investigation and monitoring data,the micro-seismicity-constraint heterogeneous EPM model and the homogeneous EPM model without the micro-seismicity-constraint were established.Based on the homogeneous EPM model,tracer test data inversion results show that the inter-well permeability is 1170 m D.Based on the micro-seismicity-constraint heterogeneous EPM model,the tracer test data inversion results show that the reservoir permeability is in the range of 2.73～10,920 m D.The high-permeability zone distributed along the NE direction of well H04 is in agreement with the spatial characteristics of low-resistance zones identified by the magnetotelluric monitoring data.Compared with the homogeneous EPM model,the micro-seismicity-constraint EPM model improves the prediction accuracy of produced tracer concentration and inter-well circulation pressure by 46.6%and 29.74%,respectively.（3）The DFN model of Habanero EGS reservoir was established based on the EPM model,and the fracture network parameters were determined by the micro-seismicity-constraint tracer test data inversion method.The inversion results show that the average fracture intensity is 0.059 m^-1,the fracture equivalent radius is in the range of 70～130m,the hydraulic aperture is in the range of 1.0～12.8 mm.Compared with the stochastic DFN model without the micro-seismicity constraint,the prediction accuracy of the produced tracer concentration and the inter-well circulation pressure is improved by67.35%and 60.44%,respectively.Compared with the micro-seismicity-constraint EPM model,the prediction accuracy of the produced tracer concentration and the inter-well circulation pressure is improved by18.99%and 51.63%,respectively.However,the computational efficiency of the DFN model is low,and a single iteration takes more than three hours.The inversion time of the fracture network parameters can be reduced to less than 10 minutes using the LSTM surrogate model integrated with the Harris inversion algorithm.In conclusion,the micro-seismicity-constraint tracer test data inversion methodology system for hydraulically stimulated fracture network characterization in the HDR can effectively improve the inversion accuracy of fracture network parameters using only the spatial and temporal distribution characteristics of induced micro-seismicity.However,the constraint of micro-seismicity source mechanism on the fracture attitude was not considered,which will introduce uncertainty to the inversion results.This methodology has low requirements for monitoring data and high inversion efficiency.It has been applied and verified using the site monitoring data of Habanero EGS,which can provide technical support and theoretical basis for production evaluation and optimization strategies for well layout and operation.It is of great significance for the sustainable development and utilization of the HDR.