Forecasting semi-dynamic response of natural gas networks to nodal gas consumptions using genetic fuzzy systems

-Semi-dynamic behavior of natural gas distribution network and nodal gas consumptions are predicted. Traditional Hardy-Cross method for analysis of the gas network is replaced with a direct mathematical solution of mass conservation equations at network nodes to yield nodal static pressures and volumetric flow rates for the coming days. After the calculation of static pressure distribution in a network for near future days, the problem of pressure drop in the network which is a serious problem in cold seasons can be managed in advance. TSK (Takagi-Sugeno-Kang) fuzzy system is used for forecasting. Structure identification of the system is carried out using CVIs (Cluster Validity Indices) and PFCM (Possibilistic Fuzzy C-Means algorithm) to determine number of rules which is also chosen such that testing error of the system does not exceed a predefined value. Premise and t-norm parameters of the TSK system are tuned by GAs (Genetic Algorithms) and their consequent parameters are adjusted using LSE (Least Square Estimate). Comparison of testing error of the TSK system for modeling benchmark data with other popular methods demonstrates its suitability for forecasting nodal gas consumptions.