Research On Daily Load Forecasting Model Of Urban Natural Gas

As a clean and efficient fossil energy,natural gas plays an important position in the world's energy use.In order to optimize China's energy structure,our government has introduced a series of policies to guide the development of natural gas industry in recent years.Thanks to the strong guidance and encouragement of national policies,the natural gas industry has developed rapidly in China.However,due to the particularity of natural gas,it needs to be stored at low temperature and high pressure,etc.The large storage of natural gas will bring some hidden dangers.Currently,it is not suitable for store in large quantities.That will provide scientific and reasonable basis to the construction of gas pipeline network and Natural gas scheduling by accurately predication of natural gas load in a future time.At the same time,it is of great practical significance to improve the operation efficiency of natural gas companies and increase the economic benefit and ensure the gas adequacy of the users.This thesis mainly research on the civil natural gas load of Datong city,Shanxi Province.By studying the natural gas load data two daily load forecasting model of natural gas was built.The main researches of thesis are as follows:(1)By reading literature and field research,several different natural gas load forecasting cycles were studied,and the data collected for short-term load forecasting of natural gas was determined.By studying the characteristics of the collected data,the data were preprocessed with the missing value filling,noise removal and data normalization.(2)The time series characteristics of natural gas load sequence are studied by means of stationarity and pure randomness test,which can give a reference to the future build time series model.By studying the influence of meteorological factors on natural gas load and the correlation between them,it provides a basis for building neural network model based on improved artificial fish swarm optimization.(3)By studying the time series characteristics of natural gas load,an ARMA prediction model based on empirical mode decomposition is built.Firstly,this model analyzes the non-stationarity of natural gas data and decomposes it with empirical mode decomposition method to obtain the finite subsequence.Then the ARMA model is built separately for the subsequence,and the final prediction results are the sum of the prediction results of each sub-series.Experiments show that this algorithm is more accurate than other comparison algorithms,and can be used to predict natural gas load in this area.(4)By studying the influence mechanism of meteorological factors on natural gas load,a back-propagation neural network forecasting model based on improved artificial fish swarm optimization is established.This model improves the artificial fish swarm algorithm and optimizes the selection of initial connection weight and node threshold of neural network.The experimental results show that the proposed algorithm can predict the natural gas load effectively,and it is better than the other two algorithms,which proves the feasibility and effectiveness of this algorithm in the prediction of natural gas load.In this thesis,two kinds of prediction algorithm is proposed from different view.These two algorithms have made good prediction results.The prediction error is controlled within 10%.In other word,it is demonstrated fully that the research method is useful with highly feasibility to predict gas load.