Statistical methods for assessing downhole casing integrity and predicting casing leaks
The present disclosure describes methods and systems, including computer-implemented methods, computer program products, and computer systems, for predicting casing leaks. One computer-implemented method includes obtaining raw electromagnetic (EM) corrosion data for a hotspot, wherein the raw EM corrosion data includes a metal thickness loss value for the hotspot; determining a corrosion parameter based on the raw EM corrosion data; generating a probability distribution curve for the corrosion parameter; generating a corrosion severity classification matrix based on the probability distribution curve, wherein the corrosion severity classification matrix includes a plurality of corrosion severity classes and cutoff values for each of the plurality of corrosion severity classes; and assigning the hotspot into one of the plurality of corrosion severity classes based on the corrosion severity classification matrix and the corrosion parameter of the hotspot.
1. A method for predicting downhole casing leaks of hotspots, comprising:
obtaining, by a data processing apparatus and from corrosion logs generated by an electromagnetic (EM) logging tool, raw EM corrosion data for a hotspot, wherein the raw EM corrosion data includes a metal thickness loss value for the hotspot;
determining, by the data processing apparatus, a corrosion parameter based on the raw EM corrosion data, wherein the corrosion parameter includes a corrosion rate that is calculated based on an original thickness value, a remaining thickness value, and a number of years;
generating, by the data processing apparatus, a probability distribution curve for the corrosion parameter;
generating, by the data processing apparatus, a corrosion severity classification matrix based on the probability distribution curve, wherein the corrosion severity classification matrix includes a plurality of corrosion severity classes and cutoff values for each of the plurality of corrosion severity classes;
assigning, by the data processing apparatus, the hotspot into one of the plurality of corrosion severity classes based on a comparison of the cutoff values for the plurality of corrosion severity classes in the corrosion severity classification matrix and the corrosion parameter of the hotspot; and
generating, by the data processing apparatus, a service alert for the hotspot based on the assigned corrosion severity class.
2. The method of claim 1 , wherein the generation of the probability distribution curve comprises:
determining whether the hotspot is a leaking hotspot or a non-leaking hotspot;
if the hotspot is a leaking hotspot, generating a probability distribution curve for the corrosion parameter based on corrosion parameters for leaking hotspots; and
if the hotspot is a non-leaking hotspot, generating a probability distribution curve for the corrosion parameter based on corrosion parameters for non-leaking hotspots.
3. The method of claim 1 , wherein the raw EM corrosion data includes metal loss values for a plurality of strings, and the corrosion parameter includes an average remaining barrier that is calculated based on a nominal thickness value for each of the plurality of strings, and a thickness loss value for each of the plurality of strings.
4. The method of claim 1 , wherein the generation of the probability distribution curve for the corrosion parameter comprises:
calculating a mean value and a standard deviation value of the corrosion parameter for leaking hotspots;
calculating a mean value and a standard deviation value of the corrosion parameter for non-leaking hotspots;
generating a probability distribution function of the leaking hotspots based on the mean value and the standard deviation value of the corrosion parameter for leaking hotspots; and
generating a probability distribution function of the non-leaking hotspots based on the mean value and the standard deviation value of the corrosion parameter for non-leaking hotspots.
5. The method of claim 1 , wherein the cutoff values for each of the plurality of corrosion severity classes is determined based on a mean value of the corrosion parameter for leaking hotspots, a standard deviation value of the corrosion parameter for leaking hotspots, a mean value of the corrosion parameter for non-leaking hotspots, and a standard deviation value of the corrosion parameter for non-leaking hotspots.
6. The method of claim 1 , wherein each of the plurality of corrosion severity classes is associated with a logging frequency for the respective corrosion severity class.