| Since the development of city fuel gas has been at the stage of the development on natural gas, the forecasting method of traditional fuel gas load would not meet the requirement for accurate prediction. It is necessary to find a new method to forecast the fuel gas load.Based on the study of fuel gas load, the actual data should adopt different methods to discover its regularity for different objects and to be verified in practice. Because of the short history of Xi'an fuel gas development, there are few required history data of natural gas load or necessary document. A gray -neural network combination forecasting method is presented to predict Xi'an natural gas load in the ten or more years. At first, the gray GM(1,1) model applied on basis of the different history data to forecast the Xi'an city's long-term natural gas load. Secondly, the gray GM(1,1) model is modified by revised parameter α method, original data processed mean method and the equal-dimension new signal method to get the forecasting values from above modified forecasting methods. Furthermore, the weighted forecasting method on gray related degree are used to get different forecasting values from above modified forecasting methods. At last, the national economy development are considered to the forecasting model by using artificial neural network, which makes the combination method more practical and reasonable.In comparision with the traditional gray method and the combination forecasting method, the gray -neural network combinatorial model obviously improves the forecasting accuracy, reduces the mean absolute error and the average relative error between actual load values and forecasting values. The combination method can also avoided the limitation of simple model, reduced the risk of the load forecasting. The comparison of actual natural gas load and forecasting load in 2005 proves the high precision of this kind of forecasting method and more adaption.
|
PDF Full Text Request