Field development studies by neuro-simulation: An effective coupling of soft and hard computing protocols

This paper establishes the principles of a novel simulation methodology that effectively brings together the specific advantages of hard and soft computing techniques. Accordingly, the work described in this paper conjoins the accurate and precise nature of the existing hard computing algorithmic approaches with the pragmatic nature of soft computing techniques. The capabilities of the proposed approach are demonstrated by the implementation of the process to various field development studies practised in the petroleum and natural gas industry. In structuring the proposed methodology, full advantages of the deterministic techniques (in terms of the accurate description of the physical phenomena) and the parallel processing power of the artificial neural network technology are taken into account to ensure the solution integrity and solution speed during the overall implementation of the process. By applying the neuro-simulation methodology to field development problems, it is shown that a substantial reduction in cost, energy and time factors of the conventional simulation approach are achieved by the pragmatic character of the artificial neural network technology. Various case studies involving different reservoirs with different flow dynamics in an increasing level of complexity are examined to illustrate the implementation of the neuro-simulation guidelines as well as to highlight the power of the proposed methodology. (C) 2000 Elsevier Science Ltd. All rights reserved.