Gas Reconciliation with Advance Error Reduction

Abstract

Pipelines are the most economically viable mode of transportation for oil and gas. Every pipeline is monitored 24×7 using meters distributed across the pipeline. Flow, temperature and pressure meters are the most common and essential for continuous and efficient operation of pipelines. Like any other instrument these meters also have uncertainty and prone to error due to irregular calibration, drift, gross error and other such events. The overall accuracy of pipeline metering increases as the distance between consecutive meters decreases. It is also affected by the placement of meters at critical locations like pipeline tapouts, tapins and consumers points. Economics do not allow pipeline operators to install beyond a certain amount of metering assets. The complexity to efficiently calculate the product in and out of a gas pipeline is more compared to a liquid pipeline. It arises due to the high compressibility of gases compared to liquids. Gas pipelines operate at much higher pressure than oil pipelines. The trapped gas inside a gas pipeline can be called line pack of that pipeline. The line pack is very sensitive to two natural factors pressure and temperature of the pipeline. Oil pipelines carry one fluid at a time. Gas pipelines on the other hand carry several gases as a mixture. Unlike oil, gas billings are calculated as the energy the gas mixture carries to the consumer. Due to the mixture, gas composition is another essential factor to accurately calculate energy of the mixture. This paper discusses the challenges of calculating various transport factors and phenomena in gas pipelines and how methods like gross error correction and machine learning can be used to increase the accuracy. The results and conclusions are derived from the applications of these methods to natural gas transportation pipeline. Some of most important conclusions obtained were Understanding the pattern of on-field meter data with ideal meter provides insights in the root cause of the problem. e.g. sudden spike in temperature leading to error in line pack.Creating digital twin of all metering assets allows faster isolation of pipeline sections having calculation errors. e.g. by monitoring the difference between field and ideal parameters.Having a central meter diagnostics system that combines the data from meters of different make and models improve the pattern recognition and error detection ability.Gross error detection isolates the meters inducing error. The feedback can be provided to the machine learning algorithms for root cause analysis. Note: This paper only covers the gross error of meters. There are methods used to reduce other meter errors namely random, limiting and systematic not covered in this paper. Readers are requested to read relevant material to understand the complete scope of errors in metering systems.