Performance test is an important part of the commissioning procedure after natural gas transmission pipeline construction. It is common that the natural gas pipeline always experience unsteady injection and withdrawal situations at different meter stations. This makes the estimation of pipeline parameters very difficult due to the uncertainty and dynamics natural of gas flow.
A Markov Chain Monte Carlo (MCMC) method is introduced in this paper to find pipeline parameters that enable the best match to the performance testing data. The parameters to be determined using this method include the specific gravity of the gas, the effective roughness of the pipeline, black powder mass concentration and average particle size. The influence of black powder on gas pipeline pressure loss is included based the experience of a commissioned gas pipeline.
The method presented in this paper consists of three major steps: data collection, data uncertainty modeling, and the MCMC simulation to determine the most likelihood parameter. Synchronized pressure, temperature, and flow rate data from compressor stations, mainline block valves, and metering stations are collected from the DCS as the basis for input and benchmarking. The variation of gas pressure, temperature, and flow rate at all meter stations are statistically analyzed to derive the probability distribution for fluctuations. In addition the probability distribution for uncertainty (and reliability) of flow meters, pressure and temperature transmitters are modeled based on published data. The specific gravity, black powder concentration and average size variation are then added with uncertainties. The final step is the estimate of the specific gravity, effective pipeline roughness, and black power concentration and sizes based on the MCMC simulations.
A real long distance gas pipeline performance test result is presented as an example to show the method is used. Instead of identifying a single parameter such as pipeline roughness the proposed method helps to provide estimate of multiple parameters based on the data. The results show better matching between the measurement data and hydraulics model prediction because of the consideration of more physical aspects.