Research On Short-term Load Forecasting Of Electrical Power System Based On Fractal Theory

Load forecasting is a significant content in terms of intelligentization and modernization of electrical power systems. Accurate short-term load forecasting (STLF) can start-up and stop of the generators economically and soundly, improve the utilization of the resources, arrange checking-repair soundly, insure the normal activities of the social production and life, improve the economical and social benefits of electrical power enterprises, optimize the safe operation of the grid and its scheduling. Hence it possesses important practical significance to the research on STLF of electrical power systems.Investigations on STLF in electrical power systems have already extend a long history. The conventional methods include regression analysis, trend extrapolation and so on. The novel ones include ANN, expert systems (ES), gray prediction, and fuzzy theory etc. The characteristics of regression analysis and trend extrapolation are described as simple structure, easy principle and rapid forecasting speed. However, the methods have the larger difficulties of initialize the model, in addition, the influence of the diverse factors has been considered. Though these new methods possess many advantages, the flaws still exist. For instance, ANN exposes some flaws such as low learning speed, and sometimes the result doesn't converge. Compared with other methods, ES does more working quantity to construct ES knowledge base. With the rapid development of electrical power systems, especially of electrical power market, the existed methods can't meet the requirement of electrical power systems. Hence, deeper investigations need be done to satisfy the increasing requirements of electrical power systems.Fractal theory is the most active branch of the non-linear science, what it concerns is non-smooth and non-differential geometry produced by complex systems. Fractal theory describes the unification between confirmation and randomicity, also order and disorder in non-linear systems. Power loads own the fractal characteristics such as self-similarity and scale-invariability, whose overall properties can be deduced out from local characteristics. With the development of electrical power market, and the implementation of price bidding, time-sharing price, and real-time price, the market will be a major factor to influence the loads. Fractal theory can overcome any market factors effectively without any assumptions, and the result will sure to be converged. Hence fractal theory is introduced into STLF in electrical power systems.Below is the major works in the thesis.(1) To apply C-C algorithm to calculate the time-delay of power loads. First, the flaws of the C-C algorithm are modified, then the revised C-C algorithm is applied to power load data. The applied investigations indicate the revised C-C algorithm is more correct. (2) To analyses the loads curve of electrical power systems, including the daily load, weekly load and yearly load. The periodic law can be found from power loads variation. Through applying fractal theory to fractal dimension and Kolmogorov entropy of power load, the results indicate that power loads possess self-similarity in terms of same space state, same time-scale, also diverse space and time-scale. This shows power loads possess fractal characteristics.(3) According to fractal theory, a method is designed for STLF based on load data in south region in Sep 2009 in this thesis. Compared with ANN, the forecasting results shows that STLF based on fractal is feasible, and possesses higher accuracy.