Abstract
For hydraulic fracturing in unconventional reservoirs, propped volume is the key to predicting total production. Microseismic analysis is frequently performed to detect fracture extension. Recently, microseismic analysis was utilized for various objectives. Though several methods of estimating proppant distribution are currently used, including tracers and core analysis, it is important to investigate whether microseismic analysis can discriminate proppant injection in order to cross-check and decrease the uncertainty of proppant distribution. To estimate the propped fracture distribution, accurate microseismic processing is important. In this study, indications of microseismic events caused by proppant injection were interpreted by accurate microseismic processing and the waveform-similarity clustering of microseismicity in the Horn River shale gas field.
We analyzed a certain hydraulic fracturing stage of zipper fracturing. In this stage, hydraulic fracturing was carried out over 7000 seconds. The first proppant injection started 1000 seconds after starting water injection and continued for 2500 seconds. Next, larger proppant was injected after termination of the first proppant injection and continued for 3500 seconds. Microseismic data was acquired using 36 downhole arrays of 3C geophones in a vertical section of two horizontal wells. Microseismic events were located by focusing P-waves, because, in this field, there is high uncertainty in picking direct S-waves. After that, we classified microseismic events using waveform similarity clustering by cross correlation between each microseismic event in the P-waves and S-waves of the each microseismic data.
Microseismic hypocenters were distributed in a bi-wing pattern from the perforation location. They were concentrated more on the southwest side than the northeast side. Sparse distribution in the northeast side could be caused by existing fractures from the neighboring well treatment. In a time-series histogram of microseismic event frequency, the events are concentrated at the start of water injection and second proppant injection. In a time-distance plot, a linear feature was detected which implies initial fracture opening. Applying the clustering procedure, two clusters were detected which include events that mainly occurred at the start of water injection. These events are located close to each other and have high waveform similarities, which implies that they have the same source mechanism, namely, initial fracture opening. Two other clusters were also detected that include events which only occurred at the start of the second proppant injection. These events were also located close to each other. From net pressure analysis, proppant screenout did not occur. Therefore, the events in the latter clusters are interpreted to have been caused by proppant injection and their hypocentres could represent proppant distribution.
Although further investigation, including fracture propagation simulation, is required and only a few events are interpreted as events caused by proppant injection, this approach should help to estimate the distribution of propped fractures and the total propped volume.
