This paper describes the development of a rod pumping expert system in which methods of emergent computation are extensively used. These methodologies include artificial neural networks (ANN) and genetic algorithms (GA). The system is implemented in order to provide assistance to the production engineer both in the problem diagnostic of operating units and the design of rod pumping facilities. The chosen computational methodologies prove to be very efficient in executing these tasks and require considerably less implementation time than other more common expert system methods.
The problem diagnostic method, implemented in the system, is based on the analysis of field dynamometer data by means of an ANN. These ANNs are computational tools, inspired on models of biological neural systems, with remarkable signal processing capabilities proven to be effective in areas such as pattern recognition and classification. The diagnostic task is achieved by using an ANN to analyze the down-hole pump cards and by means of its pattern recognition ability evaluate the condition of the pumping equipment. The ANN reacts to gross and subtle data features in the down-hole card and produces a response generalized from the "knowledge" stored in the network during a training process.
The facility design module of the expert system is implemented by means of a GA which executes an "intelligent" search in "pump configuration space" in order to produce automatically and fast several optimal design proposals compatible with the equipment restrictions and the "target" production entered by the user. Genetic algorithms (GA) are global search methods inspired in the principles of genetic evolution, which have been found to be very well suited for a multiparametric search in complex spaces. This outstanding search capability of GA is exploited in the configuration of rod pumping system designs based on the API RP 11L procedure. A run of the design procedure outputs several optimal configurations with the corresponding strokes per minute, plunger size, pumping speed, sucker rod type and operating characteristics.