Tennessee Gas Pipeline's Experience With Optimization

Tennessee Gas Pipeline's Planning Department (Planning) is responsible for identifying and implementing strategies for maximizing the return of existing pipeline assets by minimizing operating costs or increasing revenue opportunities. One such strategy involves the development of a pipeline optimization tool. In 2002 Planning developed an in-house tool that utilized an off-the-shelf optimization program. The primary drawback to this approach was the runtime required to generate a solution. TGP conducted a pilot study with Adaptive Decisions' "Adaptive Enterprise Optimization" software and based on the results of the study, elected to implement the AD Optimizer. This paper will discuss TGP's reasons for implementing an optimization tool, describe the lessons learned from that experience and outline plans for future enhancements to the optimization program. The paper will also describe TGP's reasons for selecting the AD Optimization software and how it overcomes the drawbacks of alternative optimization tools. Contrary to prevalent pipeline optimization approaches, The AD Optimizer algorithm is based on mathematical programming, not on simulation. This allows the software to consider all possible continuous values of discharge pressures, rather than discrete samples, as is done in simulation-based approaches. Furthermore, the AD optimizer allows the solution to fully utilize the available line pack, which typically results in lower fuel consumption. Finally, the AD Optimizer is built on top of a modular Adaptive Enterprise Optimization server and studio which allows the user to easily extend the optimizer with additional pipeline components, such as regulators or storage, as well as with optimization objectives, such as maximizing capacity or total profit. The AD Optimizer uses a combination of continuous nonlinear and combinatorial optimization techniques. The key unique idea is to (1) create a non-linear approximation of compressor stations off-line by pre-processing stations, (2) use the non-linear approximation to quickly figure out the optimized ON/OFF configuration, and near-optimal pressures, and then (3) refine the optimization with non-approximated equations. TGP's pilot study with the AD Optimizer demonstrated significant improvements in run time as well as solution quality over the in-house tool. The interface with the AD Optimizer is through a TGPdeveloped graphical user interface which formulates current pipeline scenarios using SCADA data and user input. Although the majority of the process is automated, each optimization is monitored and reviewed with a steady-state pipeline simulation and spreadsheet analysis. Planning operators then report the results to Gas Control. Because of operating constraints, Gas Control cannot immediately implement optimizer recommendations. Instead, gas controllers monitor the optimization results for trends to use operationally. For example, a station which was underused now operates a majority of the time as a result of continual recommendations from the optimizer. Gas controllers also have access to a historical database of previously optimized scenarios which may be used as a reference for operation. Additionally, TGP is expanding the scope of the intial project to implement station optimizations. The AD Optimizer is capable of optimizing at the station level as well as full pipelines.