Font Size: a A A

Research On The City Gas Load Forecasting

Posted on:2007-08-04Degree:DoctorType:Dissertation
Country:ChinaCandidate:Y S MiaoFull Text:PDF
GTID:1102360212970117Subject:Heating, Gas Supply, Ventilation and Air Conditioning Engineering
Abstract/Summary:PDF Full Text Request
The commercial operation of the west-east natural gas pipeline project signifies that the arrangement of the national gas pipeline network starts to change greatly. The respective city's single network system has been gradually changing to the large-scale national network system. Finally the recourse of the natural gas will become diversified , the supply will available through network, the scale of natural gas market will larger than before. Obviously, this switch will greatly change the gas supply system on the aspect of gas storage system, peak-shaving as well as the transfer and administration. Therefore, we need to grasp the characteristics and changing regular pattern of the gas load. In this way we can make load forecast of the gas active system accurately and reasonably and so we can realize the effectively operation, optimal operation and scientific administration. As mentioned above, it's a meaningful subject urgently waiting to be settled. As to the present research state of the gas load, this paper make deep research of the city gas load forecast system with a wealth of advance intelligent method. I adopt the data mining, wavelet analysis, neural network, support vector machines and multi-combined forecast theories. The main contents are as followed:Load forecast need to extract the regularity from a great number of history load variation and related elements and so we can see the correctness and detail of the history load data is quite important to the quality of all kinds of load forecast model. While the breakdown of the system and the lost of data will destroy the load regularity. The thesis paper is based on the outliers data mining theory, and it examines and locates the abnormal data in the gas history gas array by the means of k-nearest neighbors. What's more it revises the data by the means of characteristic curves.Hourly gas load fluctuates frequently and periodically. In accordance with this characteristic, this paper puts forward multiresolution wavelet network to estimate it. The hidden layer of the multiresolution wavelet network, activation function adopts the orthogonal wavelet function and scaling function. Therefore, we can separate the large-scale calculation on many layers by using...
Keywords/Search Tags:gas load forecasting, outlier data mining, wavelet nueral network, support vector machine, similar days
PDF Full Text Request
Related items