Short Term Power Prediction Of Wind Farm Based On GAVMD-SGRU Model

The power industry is the cornerstone of the development of the modern national economy.Under the national consensus to achieve the goal of carbon neutrality,the proportion of wind power in the installed capacity of the power system will continue to grow,and it will gradually become one of the important industries that promote the transformation of my country’s energy structure.However,because wind energy itself has the characteristics of strong volatility and uncertainty,in order to better integrate the wind power industry into the construction of smart grids,it is necessary to accurately predict the short-term output power of wind farms in order to reduce wind power access to the main The negative impact of the network on the stable operation of the power system.A single prediction model has a poor prediction effect when faced with nonlinear and uneven wind power data.Based on signal decomposition and deep learning,this paper constructs a short-term wind farm output power combined prediction method based on the GAVMD-SGRU model.And combined with the actual sample data for simulation,the results show that the model has a better prediction effect than the traditional single prediction model.In this paper,the statistical analysis of wind power historical data set is carried out,and the monthly and daily output characteristics of wind farm are studied in detail.Aiming at the abnormal data in wind power historical data set,the moving average method is used to preprocess,and the abnormal data that affect the accuracy of wind power prediction model is eliminated.After comparing and analyzing the respective calculation principles of the cyclic neural network model and the advantages and disadvantages in the prediction process,it is determined to select the GRU model as the basis for short-term power prediction of wind farms.Secondly,it aims at the problem of low prediction accuracy and unstable prediction of a single prediction model.Based on the integrated algorithm,a short-term power prediction method for wind farms based on Stacking fusion of multiple different GRU models is proposed.Firstly,three multi-layer GRU neural network models are built to establish the first level model,and the high-dimensional time series feature relationship is extracted deeply.Then,the training set is constructed by the prediction results of the first level model,and the second level GRU model is trained by the newly generated training set,The second level GRU model adopts a single-layer structure,which can find and correct the prediction errors in the first level model and improve the overall prediction results.Finally,the stacking fusion model with two-level model embedding is obtained.Finally,taking the measured data as an example,the practicability and feasibility of stacking GRU model are verified.Finally,in view of the difficulty in extracting the characteristic information of the original wind power data with non-linearity,this paper uses the time series decomposition method for preprocessing,which reduces the modeling difficulty of the prediction model.Aiming at the common end effect problem in VMD decomposition algorithm,this paper adopts GRU continuation method to deal with and establish GVMD model.At the same time,because the parameters K and α in the GVMD model are difficult to manually determine,this paper proposes to use the ALO optimization algorithm to optimize the parameters in the GVMD model,and compares with the PSO-VMD model to verify the feasibility of the method.Then,based on the original theory,the GAVMD-SGRU wind power combined prediction model is constructed.The model first decomposes the original wind power sequence into multiple modal components using the GAVMD model.Then the Stacking-GRU model is used to predict different components,and the prediction results of each modal component are reconstructed to obtain the predicted value of the original wind power sequence.Finally,the accuracy of the model is verified by taking historical data of the Ningxia Taiyangshan wind farm as an example.Practice shows that the model has better predictive capabilities.