Font Size: a A A

Research On Short-term Wind Power Prediction Based On Multi Neural Network Combination

Posted on:2023-09-05Degree:MasterType:Thesis
Country:ChinaCandidate:R JiaFull Text:PDF
GTID:2568306617473714Subject:Circuits and Systems
Abstract/Summary:PDF Full Text Request
The randomness and volatility of wind power bring challenges to the stable operation of power system.High-precision wind power prediction is a powerful tool to maintain the safe and stable operation of power system.With the increase of power grid stroke permeability,the prediction accuracy of single model can not meet the needs of stable operation of power system.In order to further improve the prediction accuracy of wind power,the prediction models of multiple neural networks are established by integrating convolutional neural network(CNN),long short term memory(LSTM)network and gated recurrent unit(GRU)network.Firstly,preprocess the data set.In order to avoid the abnormal data in the wind power data damaging the internal distribution law of the data and affecting the learning and training of the prediction model,the quartile method is used to clean the data twice according to the wind speed and wind power rate.The results show that after two data cleaning,the Pearson coefficient of wind speed and wind power is 0.952.Then,the missing data and abnormally cleaned data are filled based on interpolation method,which provides good quality data for the subsequent research of prediction model.Secondly,a CNN-LSTM&GRU short-term wind power prediction method based on adaptive weight is proposed.For the scenario that the new wind farm has no historical wind power data and can only be predicted by numerical weather prediction(NWP)data,in order to further improve the accuracy of wind power prediction,a short-term wind power prediction method based on CNN-LSTM&GRU multi neural network is proposed,and an adaptive weight module is proposed to select the best weight for the output of CNN-LSTM and GRU parallel prediction module,Then,a CNN-LSTM&GRU short-term wind power prediction model based on adaptive weight is constructed.The feasibility of the proposed combined prediction model is verified by using the data of a wind farm in Northwest China.The example results show that the Mean Absolute Error(MAE),Root Mean Square Error(RMSE)and determination coefficient(R-square,R2)of the proposed model are:3.395MW,2.608MW,0.959.Compared with single model and other combined models,the prediction error is smaller.Then,a combined short-term wind power prediction method of CNN&LSTM-GRU based on CEEMD-SE is proposed.Based on the deep mining of data features,a CNN&LSTM-GRU combined short-term wind power prediction method based on CEEMD-SE is proposed.Complementary Ensemble Empirical Mode Decomposition(CEEMD)and Sample Entropy(SE)were used to decompose the original wind power.It reduces the volatility and complexity of wind power data.The CNN&LSTMGRU short-term wind power prediction model is constructed by combining the advantages of CNN and LSTM network in processing local and temporal characteristics of data,and the rationality of the model are verified by using the data set above.The example results show that the MAE,RMSE and R2 indexes of the proposed model are 1.867MW,1.404MW and 0.962,respectively.Compared with single model and other combined models,the prediction error is smaller.Finally,two quantile regression methods for short-term wind power interval prediction based on multi-neural network combination is proposed.Combining the above two prediction models with quantile regression,a quantile regression short-term wind power interval prediction model based on multi neural network combination in two different application scenarios is constructed,and the short-term wind power interval prediction is studied.The example results show that under different confidence levels,the proposed quantile regression short-term wind power interval prediction model based on multi neural network combination can take into account both effectiveness and reliability.
Keywords/Search Tags:short-term wind power prediction, short and long term memory network, convolutional neural network, gated circulation unit, quantile regression
PDF Full Text Request
Related items