Abstract：
The technique of inter-well tracer testing is considered to be one of the most direct and accurate methods to identify thief zones in reservoirs. However, in heavy oil reservoirs, tracer breakthrough curves are mostly parabolic and unimodal, resulting in small differences between curves and difficulty in determining the level of thief zones. Because of the need to specify the number of clusters in advance, the Fuzzy C-means (FCM) algorithm is not suitable for the problem of unimodal tracer breakthrough curve classification. To this end, an improved FCM clustering algorithm is proposed in this paper to solve the problem
of classifying unimodal tracer breakthrough curves. First, a decision graph is introduced to determine the cluster number to avoid the blindness of FCM algorithm in selecting the cluster number. Then, the characteristic parameters of tracer curves, such as migration velocity, peak concentration and recovery rate, were selected to cluster the wells that have tracer data to determine the development levels of the thief zones. At the same time, the main factors of the development levels of thief zones are determined based on the tracer curves history matching and clustering results. Finally, the rationality of classification results is verified by their Lorenz coefficient and the application effects of foam flooding in the oilfield. The results show that the tracer migration
velocity is positively correlated with the permeability under the semi-logarithmic coordinate. The recovery rate is positively
correlated with the formation coefficient in a Cartesian coordinate system. The target oilfield has developed three levels of thief
zones. The original geological conditions of the light thief zones are homogeneous to relatively homogeneous, with a migration
velocity less than 1.65 m/d, the corresponding permeability is less than 3259 mD, and the recovery rate is less than 0.13
%
since
the formation coefficient is less than 1341 mD·m. The moderate thief zones are relatively homogeneous to heterogeneous, with a
migration velocity between 1.65 and 2.17m/d, a corresponding permeability of between 3259 mD and 8383 mD, and the recovery
rate is between 0.13
%
and 0.21
%
as the formation coefficient is between 1341 mD·m and 3194 mD·m. The severe thief zones are
heterogeneous to severely heterogeneous, with a migration velocity and permeability greater than 2.17 m/d and 8384 mD, respec
tively, and a recovery rate greater than 0.21
%
with the formation coefficient greater than 3194 mD·m. Field practice shows that
the more developed the thief zone, the better the foam flooding treatment effect. The research results have guiding significance
for profile control and water plugging in oilfields.
Key words：tracer; improved FCM clustering algorithm; breakthrough curves; thief zones classification; Lorenz coefficient
Corresponding Authors： junjian@cup.edu.cn
Cite this article：LU Chunhua, JIANG Hanqiao, LI Jie, YOU Chengcheng, CHENG Baoyang, LI Junjian. A classification method for reservoir thief zones based on an improved FCM clustering algorithm. Petroleum Science Bulletin, 2020, 04: 520-530.
