Research On Control Strategy Of Hybrid Energy Storage System Based On Attention-GRU Wind Power Output Prediction

With the rapid development of society,global environmental pollution and energy crisis have become an issue that cannot be ignored by all mankind.In recent years,the installed capacity of wind power in the world has been increasing,and the penetration rate has been increasing.Wind power generation has also caused many problems.Wind energy has strong volatility and uncertainty.Unstable wind power will bring impact to the grid during the process of grid connection,resulting in a decline in power quality and a huge challenge for grid dispatching.Therefore,it is of great significance to improve the controllability and grid connection stability of wind power generation.Wind power prediction technology and energy storage control system can effectively improve the controllability and safety of wind power integration.Among them,wind power forecasting technology has important significance for the dispatching and control of power systems and the production and maintenance of wind farms,and energy storage equipment can control the two-way flow of stored energy and effectively suppress wind power fluctuations.This thesis first studies the characteristics of power-type and energy-type energy storage equipment and the development status of energy storage technology,uses Fourier transform to analyze the historical power data of wind farms,and studies the output power frequency distribution of wind farms.The structure and working principle are analyzed and modeled.Based on the above theoretical research,this thesis combines wind power prediction and energy storage system to provide an effective control strategy for smoothing wind power output fluctuation.In the aspect of wind power forecasting,the Pearson correlation coefficient analysis is firstly carried out for various factors that may affect the power generation of wind power plants,and the input variables of the prediction model are selected.Then,the gradient disappears and the traditional neural network occurs during the back propagation training.Gradient explosion problem and poor learning effect in the encoding-decoding process,the Attention mechanism is introduced into the GRU network,and a wind power prediction model based on Attention-GRU architecture is proposed to deeply mine historical data and prediction data.Finally,compare with the other five prediction methods to verify the feasibility of the proposed prediction model.In terms of energy storage control,this thesis uses battery-super capacitor hybrid energy storage equipment to stabilize the high frequency and low frequency components of wind power according to the response characteristics of energy type and power type equipment,and proposes a hybrid energy storage predictive control strategy.The power prediction results are primary allocation and secondary correction:firstly,the adaptive variational mode decomposition is performed on the wind power prediction result,and the prediction error and the response characteristics of different energy storage devices are considered to obtain the primary distribution power that meets the grid-connected power fluctuation standard.Secondly,according to the state of charge(SOC)of the energy storage device,the distributed power is secondarily corrected by the proposed energy storage predictive control strategy.While ensuring the grid-connected power fluctuation standards required,the power of the hybrid energy storage system is rationally allocated to ensure the service life of the energy storage equipment.Finally,a comparative study of hybrid energy storage control strategies based on wind power prediction is carried out to verify the effectiveness of the proposed strategy.