Research And Application Of Linear And Nonlinear Combination Forecasting Model Based On Group Intelligence Optimization

Since the first industrial revolution in the 18th century,human consumption of energy has shown a sharp increase.Energy is divided into renewable energy and non-renewable energy,non-renewable energy such as oil,coal,etc.Renewable energy includes solar energy,tidal energy,and wind energy.The massive use of fossil energy has a huge impact on the environment,and extreme weather around the world is also frequent.Not only that,such as oil,coal and other fossil resources are distributed unevenly on a global scale,and oil prices are controlled by a few countries,resulting in an increase in the instability of the global energy market.Energy is a strategic resource of the country,and a country’s supply and demand for energy must not be subject to the international energy market.Due to the non-renewability of fossil energy and the great harm to the environment,clean and green renewable energy has become the focus of research and development by many experts and scholars.Wind energy accounts for a large proportion of renewable energy.The global wind energy is about 130 billion kilowatts,which is 10 times larger than the total amount of water energy that can be developed and utilized on the earth.This shows that the absolute amount of wind energy is very considerable.of.In addition,wind energy plays an important role in modern energy,not only has an important impact on energy dispatch,but also has an important impact on control.The wind speed has important value for the estimation of wind energy.In recent years,many scholars and scholars have been working on the wind speed prediction model.However,many wind speed prediction models do not take into account the importance of data preprocessing and parameter optimization,ignoring the limitations of poor prediction accuracy and poor stability of a single model,resulting in poor prediction results.For the above reasons,this paper proposes a combined model based on data preprocessing technology,parameter optimization and linear and nonlinear,aiming to improve the accuracy and stability of short-term prediction.The combined model prediction results can be used for wind energy scheduling and management.The wind energy can be fully utilized and the economic benefits of the wind farm are maximized.The structural framework of the proposed combined model can be divided into three parts:data preprocessing,parameter optimization,model prediction and result testing.The model proposed in the data preprocessing section uses the novel Ensemble Empirical Mode Decomposition method to denoise the experimental data.The parameter optimization part uses particle swarm optimization algorithm,bat optimization algorithm and cross-validation method to optimize the parameters of single support vector machine,BP neural network and generalized regression neural network respectively.In the model prediction and results test,the wind speed data of Shandong Penglai wind farm is used to predict the short-term wind speed.The combined model prediction results are compared with the results of single support vector machine,BP neural network,generalized regression neural network and differential autoregressive moving average model.The experimental results show that the combined model not only overcomes the defects of poor prediction accuracy and poor stability of single model,but also improves the prediction accuracy and stability,and provides a technical basis for wind energy management and scheduling.The research in this paper can be divided into four parts:The first chapter expounds the research background and research significance of wind speed prediction,the research status and comments at home and abroad,innovation points and deficiencies.The second chapter introduces the relevant theory of the model.The data preprocessing method adopted by the model is a set of empirical mode decomposition methods.The model optimization methods include particle swarm optimization algorithm,bat model optimization algorithm and cross-validation method.Related model theory includes support vector machine,BP neural network,generalized regression neural network and combined model theory.The third chapter uses the data of the wind farm unit of Penglai City,Shandong Province to conduct an empirical test on the combined model.The experimental data is denoised by the set empirical mode decomposition method,and the preprocessed data is input into the combined model to calculate and obtain the predicted result.Compare the results with the output of a single predictive model.The conclusion is that our proposed model performs optimally in terms of stability and accuracy.The fourth chapter expounds the significance of this paper’s research on wind energy forecasting,energy management,and energy dispatching.The fifth chapter is the future research ideas and research directions.The innovations of this paper are:a novel combination model is proposed,which combines the ensemble empirical mode decomposition technique,the advanced optimization algorithm and four prediction models,namely BP neural network,generalized regression neural network,support vector machine and autoregressive integrated moving average model.The proposed combined model successfully combines the advantages of the above four prediction models to further improve the accuracy and stability of wind speed prediction.Although the model proposed in this paper has made some breakthroughs in the accuracy and stability of wind speed prediction,there are still some shortcomings in this study.The shortcomings of this paper are as follows:Firstly,the proposed combination model focuses on the application of statistical models and neural network models in wind speed prediction.It is not involved in other methods of wind speed prediction,such as spatial correlation models and physical models.The second is that the prediction of wind speed in this paper is only to predict the wind speed in the future.In the future,we can also consider the research of multi-step prediction and further improve the proposed model.