Research On Short-term Power Load Forecasting Based On Chaotic Theories

With the continuous warming of power cooperation in the countries along the "One Belt and One Road",the power industry has become an important area for energy industry to expand its business.In order to ensure the continuity and stability of power supply,power dispatching accuracy and operating safety and economization,aiming at the problem of low power load forecasting accuracy,this paper analyzed a 3D jerk chaotic system with hidden attractor by using chaotic dynamics theories,and presented a prediction algorithm of chaotic time series-radial basis function(RBF)neural network based on the theories of chaotic time series analysis and RBF neural network prediction.Meanwhile,the prediction algorithm was successfully applied to the solution of real engineering problem of short-term power load prediction in two regions.Firstly,this paper summarizes the importance of short-term load forecasting for power dispatching and operational management,as well as various methods of short-term load forecasting.At the same time,the chaotic dynamics theories,the selection method of the best delay time and the best embedding dimension,the reconstruction of phase space and other chaotic time series analysis theories are introduced.Secondly,for a 3D jerk chaotic system with hidden attractor,we study the dissipative,equilibria and stability of the system,the dynamical behaviors of the system with different parameters are analyzed,such as attractor types,Lyapunov exponent and Poincare section,and through the circuit design,the correctness of the 3D jerk chaotic system is verified.Thirdly,the best delay time and the embedding dimension of the 3D jerk chaotic time series are calculated by the mutual information method and Cao algorithm respectively,and then,the phase space of the 3D jerk chaotic system is reconstructed.Meanwhile,a prediction algorithm of chaotic time series-RBF neural network is proposed based on the prediction theories of RBF neural network,and respectively applied to 3D jerk chaotic system,Lorenz system and Logistic system.It can be found that the prediction algorithm of chaotic time series-RBF neural network has the best effect for the time series of 3D jerk chaotic system compared to the other two systems,and has stable absolute error and minimum mean absolute error.Finally,taking the weekly power load of the same time in two regions as an example,chaotic identification is carried out,continuous power spectrum and largest Lyapunov exponent are calculated,and the phase space of load time series is reconstructed.Meanwhile,the prediction algorithm of chaotic time series-RBF neural network is applied to the prediction of the weekly power load of the same time in two regions.Comparing with the single RBF neural network prediction algorithm,the proposed algorithm has more stable absolute error,more reasonable error fluctuation,smaller mean absolute error.Moreover,the universality,feasibility and effectiveness were verified by numerical results.