Wind Farm Based On Improved Swarm Intelligence Algorithm Study On Short-term Power Prediction

With the destructive and non renewable characteristics of traditional energy to the environment and the current global energy crisis,renewable energy represented by scenery has attracted the attention of many scholars and public departments.Among them,thanks to human technology accumulation in wind energy development for many years.the cost of wind power development has decreased significantly,so it has strong practicability.However,the nonlinearity and volatility of wind energy pose a great challenge to the stable operation of power grid,which will lead to the grid enterprises’ restriction of wind power grid connection and wind abandonment.Therefore,the realtime and effective prediction of wind power is very important for the healthy development of wind power and the stable operation of power grid.After analyzing the current prediction methods,this paper puts forward two ideas to enhance the stability and accuracy of the model based on data decomposition framework and classification prediction.In order to obtain the information hidden in the data,a data decomposition framework based on global empirical mode decomposition and sample entropy is proposed,and it is fused with the GRU neural network optimized by the improved whale algorithm to verify its effectiveness;High latitude can better restore the real situation,but too high a dimension can lead to dimension disaster.This paper improves the calculation speed of the model by reducing the dimension of the data with too high dimension,and uses K-means clustering to divide the annual data into several groups to train the model parameters respectively.The experimental results show that KPCA-KGRU model can better fit the sequence under multi-dimensional information than the previous univariate model.(1)The principle of converting wind energy into electricity and the current research status of short-term wind power prediction technology are described in this paper.Through summarizing the literature,this paper introduces the principle of converting wind energy into electricity by wind turbines,and compares the advantages and disadvantages of different wind turbines at present.The mainstream wind power forecasting systems both in China and abroad are analyzed together with the short term wind power forecasting methods,and their own solutions are given in the rest of this paper;The correction method and evaluation index of wind power data are given.Reliability of data is the basis for correctly establishing the prediction model,and the SCADA data acquisition platform commonly used in power systems may have acquisition failures due to various reasons.In this paper,he outlier detection algorithm LOF is selected to modify the data to increase the reliability of the data.Finally,the dataset is normalized to eliminate dimensions in the dataset.In order to judge the advantages and disadvantages of the prediction model,the mainstream wind power prediction evaluation index is given in this paper.(2)A data decomposition framework based on empirical mode decomposition is proposed.Due to the nonlinear and complex characteristics of wind power data,the data should be decomposed to fully mine the hidden information of the data.The ensemble empirical mode decomposition(EEMD)decomposes the power data with large fluctuations into a few IMF components with small fluctuations,while these components retain the information of wind power.The sample entropy determines whether the IMF components can be merged by comparing the sample entropy values of the two IMF components.Inputting the processed IMF components as the input components of the prediction model one by one,and finally combining the output values to obtain a final prediction result;Improve the intelligent algorithm of whale swarm.Swarm intelligence algorithm is the result of scholars’ in-depth research on animal behavior in nature.The use of swarm intelligence algorithm can solve many practical engineering problems.Whale algorithm has been favored by many researchers because of its simple structure and few parameters.However,due to its linear convergence mode,the optimal result cannot be found.Therefore,this paper improves its convergence factor and introduces differential mutation strategy to increase the diversity of the population;unsupervised optimization of GRU neural network parameters based on improved whale algorithm.GRU neural network is a variant of RNN cyclic neural network,which has good performance in dealing with time series,but the parameters of neural network are difficult to determine.In this paper,the number of hidden layers and iteration times of GRU network are used as decision variables to improve the whale swarm algorithm.Obtaining appropriate parameters by improving the intelligent algorithm of the whale population;(3)The KPCA-KGRU classification wind power forecasting model is proposed: Drawing on the idea of similar days in load forecasting,a KPCA-KGRU forecasting model of wind power is proposed.Compared to single data,the original dataset has higher dimensionality,which will increase the processing time and reduce the usefulness of the model.Therefore,choosing KPCA algorithm reduces the data dimension while preserving the key information of the data to avoid the dimension disaster.Classify the dataset based on the KMEANS algorithm to increase the robustness of the model;Designed a wind power prediction experimental platform based on KPCA-KGRU.We developed a GUI platform based on the MATLB for the sake of improve the practicality of the model.At the same time,the platform not only connects the SCADA system,but also supports users to manually import data for power prediction.