Research Of Town Gas Forecasting Methods On Medium And Long Term Load For A Certain City

Compared with other fossil fuels such as petroleum and coal, natural gas has been widely used for its clean and efficient combustion. Thus the consumption of natural gas should grow steadily to meet the requirements of environmental protection and economic development.Natural gas recently became an important energy of the city. Due to the more development of natural gas industrial and the more sustainable growth of gas supply, the higher request of gas pipeline construction was demanded. Meanwhile, the imbalance of town gas supply and demand presentedincreasingly. From the above, load forecasting has positive practical significance because it was well known to solve aforementionedproblems.In this thesis, the representative research status about gas load forecasting and common forecasting models were analyzed and compared. According to the characters of selected city, the key factors and forecasting methods were chosen. In the medium term gas load forecasting, decomposition-combination model,ordinary GM(1, N) model and equivalentdimensions additional GM(1, N) model were carried out, and variance optimization was used to relate single model. In the long term gas load forecasting, ordinary GM(1,N) model and correction grey model based on the residual error were used.Based on these results,gas load data of sample city was subsequently used to test the feasibility of forecasting model, and evaluating the model through accuracy test and posterior check. In the medium term load forecasting, the results showed that the relative error of decomposition-combination model and equivalentdimensions additional GM(1, N) model was 6.25% and 8.34%,contrasting with the relative error of relation model was 5.84%, with optimal posterior error ratio and small error possibility. Furthermore, the relative error of ordinary GM(1,N) model was 19.64% in the long term load forecasting, contrasting with the correction grey model based on the residual error was 4.45%. In conclusion, it wasindicated that relation model was more suitable for medium and long term gas load forecasting in sample city.