| 摘要 |
A method for predicting the quality and yields of a crude oil by the application of neural networks and genetic algorithms, characterized in that it comprises the following phases:
a) determining a sufficient number of physico-chemical characteristics of an unknown crude oil;
b) supplying said physico-chemical characteristics to a neural network system previously trained, consisting of a neural grouping network and at least one neural prediction network wherein said neural networks are constructed and optimized by means of genetic algorithms;
c) applying said physico-chemical characteristics to said neural grouping network to associate said unknown crude oil with a predefined group to which a specific neural prediction network corresponds;
(d) applying said physico-chemical characteristics to said neural prediction network of said predefined group to predict yields and quality parameters of said unknown crude oil.
|
