Research And Application Of Ultra-short-term Wind Power Forecasting Method Based On Combined Model

With the over-exploitation and use of traditional fossil fuels around the world,environmental problems have become increasingly severe,and the threat of energy depletion has also intensified.In contrast,renewable energy sources(such as wind energy,etc.)can effectively make up for the shortcomings of traditional fossil fuels due to their permanent,clean and flexible characteristics.Therefore,many countries have made the development of renewable energy a key strategy.With the increase in the use of clean energy in recent years,wind power generation has been receiving more and more attention,and has become an important source of electricity in the power system.Research on wind power forecasting has also been increasingly valued by researchers.However,due to the volatility and randomness of wind power time series data,traditional single forecasting models cannot perform well in complex prediction tasks,resulting in lower prediction accuracy.This paper summarizes existing wind power time series prediction methods,and proposes a combined prediction model based on the "decomposition-reconstructionprediction-combination" idea,the CEEMDAN-IE-PSO-CM combined model,to improve the prediction accuracy of wind power data.The model uses the complete ensemble empirical mode decomposition with adaptive noise algorithm(CEEMDAN)and the increment entropy(IE)to denoise and reconstruct wind power data,and uses the temporal convolution network(TCN),gated recurrent unit(GRU)and random forest regression model(RFR)to form a single prediction module,respectively predicting high,medium and low-frequency components.Finally,the particle swarm optimization algorithm(PSO)is used to determine the optimal combination weights of each model and weighted sum to obtain the final prediction result.The validity and stability of the model are verified by experiments and analysis on wind power data sets publicly available from the Belgian power grid company.Based on the proposed combined prediction model,this paper designs and implements a wind power prediction system using technologies such as Spring Boot,My SQL and Tensor Flow.The system has three core functions: power data entry and preprocessing,wind power generation prediction and historical result inquiry.The system displays relevant information in a visual form,making it easy for users to browse wind power historical information and prediction results in an intuitive way.To sum up,this paper makes up for the deficiency of single forecasting model in wind power forecasting by studying the combined forecasting method to improve the accuracy of wind power forecasting,and a wind power prediction system with high prediction accuracy and strong applicability has been designed and implemented,which can contribute to better decision-making by power workers on power transfer,improving the stability and reliability of the power grid system,improving the energy structure,and reducing the use of traditional fossil energy.