Electricity Load Forecasting Based On Chaotic Neural Network

Short-term electric power load forecasting is the basis work for power system optimization running. It is very important for the electric power system stability, economic and safety running, and it is also an important work in the scheduling and load management. There is a great significance to find an effective load forecasting method to improve the prediction precision. Short-term electric load can be influenced by the weather, holidays, policies and any other factors, so the load performance shows extremely complex features which are difficult to accurately predict under various factors. Along with the development of nonlinear theory, especially the development of chaos theory, it is able to make a satisfactory short-term load forecasting result without considering various factors. Chaos is a new subject, and the neural network is a kind of intelligent technology which has a strong ability to deal with nonlinear problems, so this dissertation combines chaotic characteristics and neural network to show more complex dynamics characteristics. This dissertation researches the short-term electric power load forecasting based on chaotic neural network.Firstly, this dissertation discusses the significance and the basic principle of the electric power load forecasting, introduces the methods of electric power load forecasting and the researching situation at home and abroad, describes the relevant content of the chaotic dynamics, investigates the phase space parameters effect on the quality of reconstructed space, and the different ways which determine the number of phase space embedding dimension and delay time based on the reconstructed phase space theory. This dissertation researches identification methods for the properties of chaotic time series, which establishes the foundation for the chaotic analysis of power system.Secondly, this dissertation processes the reconstruction phase point, chooses advantaged points as samples of neural network to train. For the deficiencies that still exist in the chaotic neural network model, this dissertation proposes a method that enlarge the role of chaotic signal factors, and makes full use of chaotic characteristics to enhance prediction accuracy.Finally, this dissertation establishes a prediction model based on chaotic neural network, analyzes the working mechanism of chaos factors, and applies this model to predict electric power load. Through analyzing and evaluating the prediction results, it proves that this method is feasible.