System and method for inferring geological classes
A system for inferring geological classes from oilfield well input data is described using a neural network for inferring class probabilities and class sequencing knowledge and optimising the class probabilities according to the sequencing knowledge.
1. A system for inferring geological classes from downhole log data comprising a neural network for inferring class probabilities, characterized in that said system further comprises means for integrating class sequencing knowledge and optimising said class probabilities according to said sequencing knowledge, and storage for said inferred geological classes to establish a relationship between the inferred geological classes and the downhole log data, wherein said geological classes comprise one of lithology, rock type and petrophysical properties.
2. The system of claim 1 , wherein the means for integrating class sequencing knowledge and optimising said class probabilities according to said sequencing knowledge comprises a hidden Markov model.
3. An automated system for inferring geological classes from downhole log data, comprising a data input vector, a neural network trained to infer from said input vector a class sequence or class probability vector, and a modifier for correcting said class sequence or class probability vector using prior knowledge of class sequence or class probability, and storage for said inferred geological classes to establish a relationship between the inferred geological classes and the downhole log data, wherein said geological classes comprise one of lithology, rock type and petrophysical properties.
4. An automated system of claim 3 , wherein the modifier uses the prior knowledge of class probability distribution and class transition probability.
5. An automated system of claim 3 , wherein the modifier includes a Viterbi sequence optimisation.
6. An automated system of claim 3 , wherein the modifier includes a Bayesian based probability calculator.
7. An automated system of claim 3 , wherein the modifier includes a Bayesian based probability calculator and a Viterbi sequence optimisation.
8. A method for inferring geological classes from downhole log data, comprising the following steps:
inferring class probabilities with a neural network;
integrating class sequencing knowledge and optimising said class probabilities according to said sequencing knowledge; and
storing said inferred geological classes to establish a relationship between the inferred geological classes and the downhole log data, wherein said geological classes comprise one of lithology, rock type and petrophysical properties.
9. The method of claim 8 , wherein the integrating class sequencing knowledge and optimising said class probabilities according to said sequencing knowledge is achieved according to a hidden Markov model.
10. A method for inferring geological classes from downhole log data, comprising the steps of generating a data input based on said well input data; using a neural network to generate a class sequence or class probability vector inferred from said input; correcting said class sequence or class probability vector using prior knowledge of class sequence or class probability; and storing said inferred geological classes to establish a relationship between the inferred geological classes and the downhole log data, wherein said geological classes comprise one of lithology, rock type and petrophysical properties.
11. The method of claim 10 , wherein prior knowledge of class probability distribution and class transition probability is used to correct the class sequence or class probability vector.
12. The method of claim 10 , wherein the correction includes a Viterbi sequence optimisation.
13. The method of claim 10 , wherein the correction includes a Bayesian based probability calculation.
14. The method of claim 10 , wherein the correction includes a Bayesian based probability calculation and a Viterbi sequence optimisation.