Research On Short-term Wind Power Forecasting Based On ICEEMDAN And SSA-ELM

Realizing the rapid transformation of energy and building a low-carbon,clean,costeffective,and modern energy system are my country’s core goals in the new round of energy revolution.The efficient development and efficient use of new energy are its important features and tasks.By 2050,the global population is expected to increase to 9.6 billion.As the population increases,the global energy demand is also increasing.The consumption of a large amount of primary energy will generate a large number of greenhouse gases,which is the main cause of global warming.For this reason,people are paying more and more attention to the development of clean energy.Wind power is the world’s second-largest source of renewable energy power generation,and it will become the main trend of future development.The countries with the strongest wind power generation capacity are China,the United States,Germany,India,and Spain.However,due to the randomness and intermittency of wind energy resources,the power of wind power has great volatility.Therefore,accurate prediction technology is very important for wind power generation.Accurate wind power prediction can provide dispatch reference when wind power is connected to the grid,and provide an effective reference from the balance of power generation,transmission,and power consumption of the grid;it is conducive to effective formulation of day-ahead power generation plans,reducing wind farm operating costs,and at the same time It can also effectively reduce the random fluctuation of wind power and improve the capacity of wind power absorption.Based on the strategy of "decomposition-reconstruction-feature selection-predictionintegration",this paper proposes a complete integrated empirical mode decomposition(ICEEMDAN)and sparrow optimization algorithm(SSA)optimization extreme learning machine based on improved adaptive noise(ELM)combined forecasting model for shortterm wind power forecasting research.First,judge and filter the data collected by wind farms,and perform statistical tests,nonlinear and non-stationarity tests on the filtered data,and innovatively apply the ICEEMDAN model to the preprocessing of wind power data;The decomposed components are reconstructed according to the complexity of the sample entropy(SE)algorithm;then the chaos theory is applied to the characteristic analysis of wind power components,combined with meteorological factors and phase space reconstruction(PSR)to select prediction The input and output variables of the model;finally,the extreme learning machine(ELM)combination model optimized by Sparrow Algorithm(SSA)is used to predict each component one by one,and the final prediction result is obtained by accumulation and integration.Through empirical analysis,the selection of comparative models verifies the proposal of this article The effectiveness of the combined model.With the advancement of the global energy transition,as well as the introduction of carbon peaks and carbon-neutral goals,in future development,the use of wind energy has become the main development trend.Therefore,this article is based on providing effective means for the development of wind power grid-connected and established the short-term combined forecasting model of wind power based on the strategy of "decompositionreconstruction-feature selection-prediction-integration" has certain practical significance.