Different fields and wells produce natural gas of different composition, which directly influences the value of the gas. This gas is then sent to customers depending on the individual specification. Gases from different origins are sometimes blended to make up for the minimum rate requirement. This study presents a mathematical approach, linear programming, to process big data and generate an optimized route that solves a rate allocation problem keeping in place various operational constraints.
A base model was created based on linear programming algorithm for the proof of concept. As in industry, multiple sets of constraints were applied to the model. These constraints include, for instance, maximum allowable carbon dioxide concentration of the total blend, minimum amount of gas, or target ethane concentration for various purposes. Another feature that was included in the model was Blend Specification Requirements. This feature navigates the solver to which result is more favourable and would provide higher profit. The final goal of the solver is to provide a scenario that complies with all the constraints providing the best revenue.
The results of the multiple test optimizations, generally, showed close agreement with predictions made prior the tests. Once the model was validated, more complex scenarios were evaluated. Here, the model-generated results showed completely different from the expected. This error in predictions is due to the nature of the problem i.e., rate allocation planning, where the number of considered variables directly influences the complexity of the decision, making it impossible for basic experience-based predictions. The soft-constraint Blend Specification Requirement principles proved to function and direct the solver towards the higher profitable arrangements, while complying to the hard-constraints. Soft and hard terms here relate to the level of the obligement that must be performed by the solver. Overall, the model succeeded in optimizing the relationship between the economic values of different natural gas compositions and the operational and blend constraints, identifying the most profitable gas distribution plan.
It is no longer practical to rely solely on the experience of individuals when dealing with complex rate allocation operations. This study presents a simple, reliable and elegant method to build computer-based optimisation models. Connected to an online stream of Big Data, the models can potentially contribute to oil and gas pipeline rate allocation operations, making decisions fit-for-purpose, cost effective and reliable.