Wednesday, May 23, 2018

Science & Technology: Algorithms Pick Out Hidden Signals that Could Boost Geothermal Energy Production

Machine learning reveals cyclic changes in seismic source spectra in Geysers geothermal field (Science Advances)


The earthquake rupture process comprises complex interactions of stress, fracture, and frictional properties. New machine learning methods demonstrate great potential to reveal patterns in time-dependent spectral properties of seismic signals and enable identification of changes in faulting processes. 

Clustering of 46,000 earthquakes of 0.3 < ML < 1.5 from the Geysers geothermal field (CA) yields groupings that have no reservoir-scale spatial patterns but clear temporal patterns. Events with similar spectral properties repeat on annual cycles within each cluster and track changes in the water injection rates into the Geysers reservoir, indicating that changes in acoustic properties and faulting processes accompany changes in thermomechanical state. 

The methods open new means to identify and characterize subtle changes in seismic source properties, with applications to tectonic and geothermal seismicity.


Machine learning reveals cyclic changes in seismic source spectra in Geysers geothermal field, by Benjamin K. Holtzman, Arthur Paté, John Paisley, Felix Waldhauser and Douglas Repetto. Science Advances  23 May 2018: Vol. 4, no. 5, eaao2929 DOI: 10.1126/sciadv.aao2929