Research On Short-term Wind Speed Forecasting Based On Chaotic Theory And Improved Artificial Neural Network

Renewable energy,which presents the advantages of clean,pollution-free and renewable,has risen rapidly and become a hot spot in the world with the coming of 21 st century.Developing and utilizing the renewable energy has become the key to deal with the energy crisis,improve the environment and realize sustainable development.Wind power,with its irreplaceable advantages and mature development technology,stands out among many new energy sources and develops rapidly.However,in the actual development process,the inherent characteristic of wind power such as randomness,intermittency and uncontrollability affects the grid connection of wind power and endangers the safe and stable operation of power grid.Therefore,short-term wind speed prediction is an indispensable part in the development and utilization of wind energy.An accurate wind speed prediction can not only assist the power system to carry out timely and effective economic dispatch,but also help to mitigate the adverse effects of wind power grid-connection on power system and ensure reliable operation of power system.On the basis of chaotic theory and artificial intelligence neural network,this paper studied the short-term wind speed prediction by constructing FEEMD-SE-PSR-MFO-ELM forecasting model.Among which,the fast ensemble empirical mode decomposition(FEEMD)is employed for time series decomposition to improve the prediction accuracy.On the premise of retaining data information,the stabilization of time series is realized.Meanwhile,the calculation of sample entropy about the result of FEEMD is the gist of sequence reconstruction,which helps reduce the computation and highlight the correlation.The fast training speed and good learning performance is the advantage of extreme learning machine,but the characteristic of setting weight and threshold is easy to cause the fluctuation of prediction precision.In light of this,the moth flame optimization is introduced to optimize the parameter of ELM to enhance the performance of prediction.Furthermore,this paper illustrates the chaotic characteristic of wind speed time series and the phase space reconstruction is employed into the short-term wind speed forecasting.At last,taking the wind speed data of Inner Mongolia wind farm as sample,this paper experienced the proposed FEEMD-SE-PSR-MFO-ELM and verified its feasibility and effectiveness.The results demonstrated that the proposed hybrid model performs a higher accuracy and can meet the actual demand of wind farms.With the energy transform step by step,the fossil energy system will eventually be replaced by the low-carbon energy system,and the large-scale development and utilization of wind power must be the trend of energy development in the 21 st century.Thus,this paper,which bases on the problem of wind power grid connection and builds a hybrid model for the short-term wind speed forecasting expresses a certain reference value and practical significance.