Research On Short-term Wind Speed And Wind Power Forecasting Based On Wind Farm Historical Data

Nowadays,remarkable achievements have been made in the development and utilization of renewable energy.Among them,wind energy has also entered a high-speed development stage.However,the large-scale wind energy grid connection is bound to bring great challenges to the safe and stable operation of the power system.Wind speed and wind power forecasting is an effective way to reduce the cost of wind energy grid connection and ensure power quality.However,due to the effects of intermittent and random wind energy on the prediction results,how to achieve more accurate forecasting results is an important task in current research.In order to improve the accuracy of wind speed and wind power forecasting,and implement efficient grid dispatching,this thesis is mainly researching the short-term wind speed and wind power forecasting based on historical data from wind farms.The main research contents are as follows:1.To ensure the reliability and integrity of the data,data pretreatment is performed.Firstly,a combination of isolation forest and multiple imputation by chained equations was used to perform outliers detection and missing values interpolation on the original data.Secondly,considering the short sampling interval of historical data,the large amount of historical data,and the need for research,the data was integrated by adjusting the timing interval.Finally,the min-max normalization method was used to process the data.The results show that the detection of abnormal values are basically complete,the repair of abnormal values and the interpolation of missing values significantly improve the integrity and reliability of the data,the data integration effectively reduces the waste of computing resources,and the normalization process eliminates the impact of data magnitude and dimensions on the forecasting process.2.In order to improve the accurate forecasting of the random forest algorithm in the short-term wind speed forecasting,a combined forecasting method that an improved fruit fly optimization algorithm to optimize the parameters of the random forest regression model is proposed.Firstly,an adaptive update of the search step size was implemented in the fruit fly optimization algorithm to enhance the global optimization and local exploration of the algorithm’s ability.Secondly,the qualitative analysis of the influence of the main parameters in the random forest regression model on the forecasting results to determine the optimization parameters.Finally,the improved fruit fly optimization algorithm was used to iteratively optimize the model parameters,and the determined parameter values were input to the forecasting model to complete the short-term wind speed forecasting.Experiments show that the improved fruit fly optimization algorithm has better convergence accuracy,and the forecasting results of the combined method has the best fitting degree with the actual wind speed,and can meet the higher accuracy forecasing requirements.3.To solve the problem that the fluctuation characteristics and timing characteristics of wind power on the forecasting results,a short-term wind power forecasting method based on variational mode decomposition and long short-term memory network is proposed.Firstly,the wind power signal was decomposed into different band-limited intrinsic mode functions through variational mode decomposition to complete the noise reduction decomposition of the power data.Next,Dropout was added to long short-term memory network model for preventing overfitting,and Adam was selected to adjust the network parameters by comparing the gradient descent optimization algorithm.Lastly,the forecasting results of each mode components were obtained through long short-term memory network model,and the short-term wind power forecasing was accomplished by summing and reconstructing the forecasting results of each mode components.Experimental results show that variational mode decomposition can effectively extract the characteristics of wind power data and deal with signal fluctuations,and the combined method improves the the accuracy of forecasting results.