Pipelines that have intermittent flow can often be a difficulty for a leak detection system to manage reliably without causing issues with sensitivity, often requiring significant tuning and tweaking (Shawn Learn, 2015). Some key issues that can drive the modeling errors are (Ryan Dolan, 2017):
1. Pipeline Transients
2. Thermal Fronts (step changes in fluid injection temperature)
3. Heat Transfer and Thermodynamic Uncertainties
4. Slack Estimation and Thermal Expansion Uncertainties (shut-in pipelines)
5. Flow Meter Errors
The goal of the authors of this paper was to develop a highly reliable and robust Leak Detection System (LDS) that could complement our existing RTTM LDS. A simplified linepack estimation methodology was developed that could feed an existing MVB system for accurate pipeline inventory estimates. The model needed to use different base assumptions than the existing RTTM system, deal with the key issues mentioned above and to be easy to support and maintain.
For pipeline transients, it was determined that since the pipelines in question have relatively small pipe volume between flow meters, modeling of the transient pressures was not required (Morgan Henrie, 2016). A linear approximation of pressures adequately estimates the effect of pressure changes on linepack. In addition, any induced error was short lived (less than one minute).
It was decided that the simplest model that addressed the issue of thermal fronts was a model that uses a Lagrangian frame of reference to track the fluid elements along the pipeline. The product properties of the individual fluid elements (bulk modulus and thermal expansion coefficient) are determined by the application of the API MPMS standard to align with the flow meter assumptions for correction to standard conditions. These same correlations are used to calculate the volume that each of the fluid elements take up at any given time.
In order to deal with large Heat Transfer uncertainties, it was best if the model stay simplistic (since the inputs for a detailed model were not knowable). A simplified Newtonian Cooling heat transfer model was developed. This Newtonian Cooling model has factors which are continuously tuned using field temperature readings.
In addition, the Newtonian Cooling factors are also tuned based on previous pipeline start-ups. Once a successful pipeline start-up is completed the resultant slack estimate (calculated from the meter imbalance) is used to tune the Newtonian Cooling model (Holman, 2002). This helps to correct errors in the thermal assumptions.
The model has been developed and is currently in use on one of our pipelines. The online tuning mechanism has resulted in a very sensitive system where the cause of all false alarms to date have been identified and removed. There is a project in place to implement this system as a redundant form of leak detection across all our pipelines. A form of flow meter tuning is under early phase development and will be incorporated in order to further increase the accuracy and reliability of the system.