Research On Wind Speed Combination Forecast Based On Improved Intelligent Algorithm

As one of the most potential renewable energy sources,wind energy has attracted extensive attention from all countries.However,due to the huge fluctuation and uncertainty of natural wind,wind power will also fluctuate,which will seriously affect the reliability of wind power system and bring challenges to large-scale grid connection of wind power.Therefore,wind speed prediction is of great significance to ensure the safety and stability of wind power generation system.The uncertainty and instability of wind power bring difficulties to wind energy development,wind farm grid connection and power system stability.Therefore,accurate prediction of wind power is very important.It will help promote the development of wind energy prediction,help wind farms formulate wind power regulation strategies,and further promote the construction of green energy structure.How to reduce wind power prediction error and improve wind power prediction accuracy has become an urgent problem to be solved.Compared with traditional machine learning methods,deep learning has obvious advantages in massive data processing and nonstationary space-time prediction.It has great application potential in all technical links of wind speed simulation and prediction,and provides technical support for faster,more efficient and more accurate wind speed prediction.According to the characteristics of neural network and support vector machine prediction time series,three short-term prediction systems for accurately predicting wind speed are designed in this paper1)Wind energy prediction scheme based on hybrid mode decomposition and improved long-term and short-term memory neural network.Firstly,the improved mixed mode decomposition is used to decompose the wind speed data into trend part and fluctuation part.Wavelet analysis is used to decompose the trend part and fluctuation part.Analyze the stationarity of the decomposed data.The long-term and short-term memory neural network optimized by the improved particle swarm optimization algorithm is used to train the non-stationary sequence and noise sequence,and the autoregressive moving average model is used to train the stationary sequence.Finally,the final prediction result is obtained by reconstruction.2)A combined prediction scheme of multi factor wind speed prediction based on variational modal decomposition and improved least squares support vector machine.Fast filtering algorithm and variational modal decomposition are used to extract the data characteristics of wind resources,and the least squares support vector machine optimized by Drosophila optimization algorithm is used to predict the short-term wind speed.Finally,the wind speed range is estimated according to the point prediction results and prediction errors.3)A combined wind speed prediction scheme based on ensemble empirical mode decomposition and improved support vector regression machine.The ensemble empirical mode decomposition is used to decompose the wind speed data,and the support vector regression machine improved by sparrow search algorithm is used to predict the decomposed wind speed data.Finally,the prediction value is reconstructed to get the final prediction result.