An approach for predicting geothermal reservoirs distribution using wavelet transform and self-organizing neural network: a case study of radon and CSAMT data from Northern Jinan, China

Preliminary surveys found that Northern Jinan is rich in geothermal resources. To improve the prediction accuracy, we constructed an approach for predicting geothermal reservoir distribution using wavelet transform and self-organizing neural network (SOM). First, radon measurement and controlled-source audio-frequency magnetotellurics (CSAMT) field construction design was carried out, then the data collection, processing, and inversion were performed sequentially to obtain the radon and CSAMT data of each survey line. Next, wavelet transform was used to decompose the radon measurement data into different spatial and scale components and extract the low-frequency radon anomaly data of the target layer. Subsequently, the radon anomaly and resistivity data of the input sample dataset were input into the SOM for learning, which provided the classification results of the geothermal reservoirs. Finally, we assessed the SOM classification results and predicted favorable exploration areas depending on the drilling information and fault characteristics. Effective implementation contributes substantially towards understanding and developing geothermal resources. Article highlights A joint inversion method using SOM with CSAMT and radon data. Wavelet transform extracts radon anomalies in the deep target layer. Prediction of geothermal reservoirs distribution using SOM.