Modeling energy flow in natural gas networks using time series disaggregation and fuzzy systems tuned by particle swarm optimization

Natural gas is widely used in industrial, residential, and commercial sectors and is delivered to consumption nodes via gas distribution networks. For efficient management and utilization of this nonrenewable energy resource, High Frequency (HF) response of gas networks to nodal consumption is needed that requires solution of the network governing equations. This high frequency response could either be measured with expensive high technology hardware at each consumption node or alternatively calculated from Low Frequency (LF) data collected by inexpensive low technology gas meters, which are usually preferred. Solution of the governing equations itself requires nodal gas consumption that is recorded by LF gas meters installed at consumption nodes. The recording frequency differs from one meter to another. Gas companies use these LF meter data just for billing. This paper presents a methodology for HF study of gas networks response to nodal consumption using these LF data. A Time Series Disaggregation (TSD) method is formulated that disaggregates the LF meter readings to HF gas consumption all with the same frequency by which the network governing equations are solved. HF gas consumption of each node is employed to train Takagi–Sugeno–Kang (TSK) fuzzy system to forecast HF consumption of that node in forthcoming days. The governing equations are then solved using this forecasted nodal consumption to predict the gas network behavior in the days ahead. This enables gas companies to recognize areas of the network with high pressure drop in cold days and manage the network accordingly. The paper also presents two techniques to prevent pressure drop in the areas of the network with high gas consumption. The proposed methods are applied to a gas network with 4258 customers using available LF data recorded during three years.