Research On Static And Dynamic Simulation And Load Forecasting Of Natural Gas System

This thesis is concerned with two problems: static and dynamic simulation for natural gas pipeline network and short-time natural gas load forecasting.Static and dynamic simulation for natural gas pipe network and natural gas load forecasting is an important work of natural gas management system. Precise static and dynamic simulation and load forecasting will have an significant effect on production planning, optimizing attemper and safety analysis of natural gas network system, and will directly affects its economic benefits. Based on historical data of Shaanxi natural gas network system, static and dynamic simulation and load forecasting of natural gas system has been researched in this thesis.The thesis firstly made a research on waterpower calculation arithmetic of simple pipeline, and research results were used in other complicated pipeline system such as parallel pipeline, annular pipeline. By practical simulation, the results show that the arithmetic has a higher precise in static simulation.For natural gas pipe network dynamic simulation, the thesis adopts characteristic method to discuss. Based on theoretic calculation, analyzing practical working phenomena of the network, the thesis improved the arithmetic. Tested by waterpower calculation simulation, the simulation results show that the arithmetic meet the requirement of practical project.For short-time natural gas load forecasting. Based on analyzing tech situation at home and abroad, considering all kinds of factors which will have influence on load changes, a hybrid approach combined the Self-organizing Feature Map (SOFM) neural network with Multilayer Perceptron (MLP) is presented, and short-time load forecasting model is established. To make the prediction values with independence of the general trend, which is changed from year to year, the load data are transformed by profiles, mean value, and variance. SOFM is used for the prediction of profiles and MLP networks for prediction of daily mean and daily variance. At a result, load forecasting for 24 hours in a day can be gotten. It shows the validity of model by practiced simulations.