Natural Gas Pipe Network Optimization

Natural gas pipeline network, demanding gigantic investment and operating costs, is one of the most important infrastructures in the natural gas industry. After analyzing the domestic and foreign achievements of optimized operation technology for natural gas pipeline network, this thesis does further research in the optimized operation for natural gas pipeline network. The aim of our research is to increase the economic, technologic and social benefit of the design enterprise.Through the analysis of substantial research information, this thesis believes that the load forecast of natural gas pipeline network system is the basis for the optimized operation and the fundamental basis for developing operational program. Therefore, a large number of scholars research in the gas pipeline network load, and the study shows that the characteristics are the time series of sexual, degenerative, random and nonlinear; it is undermined by multiple factors and complex nonlinear relations. The artificial neural network (ANN) is able to get the appropriate parameters by learning, and it is used to shine upon arbitrary complex nonlinear relationship. Therefore, it is very applicable to short-term natural gas pipe network load forecast. But it easily falls into the smallest local. To overcome this shortcoming and improve forecast accuracy, this thesis use genetic algorithm (GA) replacing traditional learning algorithms to optimize the initial value. It can fit a better position search space in the solution space, and then use BP algorithm search for optimal solution in the small space. Calculation cases demonstrate that the genetic neural network forecasting fractional error can be dropped down to 2.5%.The optimal mathematical models of natural gas pipeline network operation are nonlinear planning, complicated by the conditions, so its solution size is sweeping. The traditional search methods for this kind of model are grid method and complex method. Grid method is a kind of method of exhaustion, which calculation workload is too tremendous. The deficiency of complex method is that its convergent speed is very slow. To get a better peak, it always shrinks to the center repetitiously; its workload is also too tremendous. The particle swarm optimization (PSO) is an optimized technique based on swarm intelligence (SI) and computer science, which doesn't need the differential coefficient of the function, suitable to solve the large-scale and complicated problems. So the PSO provides a new measure to solve the optimal mathematical models of natural gas pipeline network operation.In order to use the PSO to solve the models and overcome its shortcoming, that is, it is difficult to deal with equation restriction; this thesis adopts self-adapting penalty function, which can dynamically adjust penalty function according to the extent of overstepped the bounds in the calculation. This method improves the arithmetic's astringency and the precision of its solution. Shrink factor is adopted to restrict the particle's speed and position.