Research On Wind Power Prediction Based On Sparrow Search Algorithm And Gated Recurrent Unit Neural Network

With the rapid consumption of traditional fossil energy and the rapid deterioration of the environment,more and more countries have begun to promote the research and development of renewable energy.As a clean,low-cost energy source with huge reserves,wind energy has become a research hotspot in the academic community.However,wind energy also has the disadvantages of intermittency,volatility and instability,which brings great difficulties to wind power generation.Accurate forecasting of wind power is therefore indispensable.Existing wind farms are usually composed of multiple wind turbines,and by predicting the power of a single wind turbine,the output power of the entire wind farm can be calculated,thus providing an important reference for the absorption of wind power.It is in this research context that this thesis proposes a wind power prediction model based on sparrow search algorithm and gated recurrent unit neural network for wind turbines.The research content of this thesis is as follows:(1)Based on the idea of combined model,this thesis proposes a wind power prediction model(SSA-Combination)based on sparrow search algorithm and gated recurrent unit neural network.The model proposed in this thesis is composed of three sub-models: SSA-CNN-GRU,SSA-Attention-GRU and SSA-AR.The SSA-CNNGRU submodel is constructed based on convolutional neural network and gated recurrent unit neural network.The SSA-Attention-GRU submodel is constructed based on the external attention algorithm and the gated recurrent unit neural network.The SSA-AR submodel is constructed based on the autoregressive algorithm.In this combined model,three sub-models predict future wind power separately,and then the three predictions are added with a certain weight to obtain the global prediction result.(2)In order to further optimize the prediction performance of the model,this thesis introduces the sparrow search algorithm to optimize the hyperparameters of SSA-CNN-GRU,SSA-Attention-GRU and SSA-AR submodels to ensure the best prediction performance of each submodel.At the same time,the sparrow search algorithm is used to optimize the combined weights of the combined model,which further improves the prediction ability of the combined model.(3)In order to evaluate the prediction performance of the model,this thesis designs a single-step prediction experiment and a multi-step prediction experiment based on the real wind power dataset,and comprehensively tests the combined model proposed in this thesis on this basis.In various experiments,the models proposed in this thesis show good prediction accuracy and perform better than other control models participating in the experiment.The results of various experiments verify the rationality of the model design and the superiority of the performance proposed in this thesis.(4)Combined with the research results of wind power prediction model,this thesis constructs a wind power prediction system based on Spring Boot framework,microservice architecture and My SQL database.The system integrates data preprocessing,model training,data prediction and historical prediction record viewing,showing stable and reliable performance in software testing.