Font Size: a A A

Research On Short-Term Load Forecasting Model And Method Based On Chaotic Characteristic Of Electric Load Time Series

Posted on:2006-08-04Degree:DoctorType:Dissertation
Country:ChinaCandidate:S L LeiFull Text:PDF
GTID:1102360155472599Subject:Electrical engineering
Abstract/Summary:PDF Full Text Request
Electric short-term load forecasting(STLF) is an important and integral component in the operation of any electric utility whose accuracy directly influence power system's security, profit and quality. Power system is a strong nonlinear system, and appears as chaos behavior. In this paper, studies to electric load time series shows that the time series is not random series but chaotic ones, by calculating the chaotic character exponents: the largest Lyapunov exponent and correlation dimension. So it gives us a new acquaintance to the complexity of load time series, and the short-term electric load could be forecasted through chaos theory. Based on phase space reconstruction of chaotic time series, the short-term load forecasting approach on Adaptive Neural Fuzzy Inference System (ANFIS) is presented. It may adequately reflect the varying rule of the data itself, in order to improve the accuracy of load forecasting. Based on local linear prediction model of chaotic time series, short-term load forecasting method on multi-embedding dimension is presented. In the phase space reconstruction of chaotic time series, proper choice of embedding dimension and delay time play an important role in the attractor size and similar degree between reconstruction phase space and the original dynamic system. Different reconstruction parameters of the time series might be obtained by different approach. Different forecasting results are given by the different parameters. In order to improve the efficiency and practicality of load forecasting, the forecasting load is given by weight average for the results of different embedding dimension. Optimal choice method of the nearest neighboring points and adding weight one-rank local region method is introduced on the nearest neighboring forecasting method of chaotic time series. In chaotic local forecasting method, proper choice of the nearest neighboring points directly affect the forecasting results. In order to improve the forecasting model, the nearest neighboring points are selected by correlation degree between forecasting point and its neighboring points meeting the Euclidean distance. At the same, adding weight one-rank local forecasting model is set up in order to counter the influence of different neighboring points. A practical example expresses that the method is effective and feasible. To counter the influence of different factors on short-term load, a model of multivariate time series for forecasting the short-term load is set up. Based on the phase space reconstruction method of univariate time series, phase space of multivariate time series is reconstructed, and then global and local forecasting model of multivariate time series for forecasting the short-term load is set up. It is simple to be applied. A model of short-term load forecasting under the environment of power market is presented. In power market some character of electric load would be changed to a great degree due to real-time price and usage pattern variability of the consumers. So the present load forecasting methods are challenged. According to the above problem, multivariate time series forecasting method and modern intelligent combination forecasting method are discussed on real-time price. The latter first draws on the nonlinear approaching capacity of the Radial Basis Function(RBF) neural network to forecast the load on the prediction day which takes no account of the factor of electric price, and then, based on the recent changes of real-time price, uses the ANFIS system to modify the results of load forecasting obtained by using the RBF network so as to improve the forecasting accuracy and overcome the defect of the RBF network in price-sensitive environment and objectively reflect the relation between real-time price variability and electric load.
Keywords/Search Tags:STLF, chaos, multivariate time series, RBF network, ANFIS, Lyapunov exponent
PDF Full Text Request
Related items