Research On Short-term Output Power Prediction Of Wind Power Station

Due to the intermittent,random and uncontrollable nature of wind power,it brings potential safety hazards to the stable operation of the power grid.To fully utilize effective wind energy resources,predicting wind speed and wind power is a prerequisite for grid-connected operation of large-scale wind farms.Accurate wind power prediction is of great significance to the power system dispatching department to formulate the wind power grid connection plan,ensure power quality,and ensure the stable operation of the power grid.The work of this paper is mainly divided into two parts: by studying the characteristics of wind speed and wind power,and comprehensively analyzing the relevant factors that affect the change of wind power output power,we can know that the size of wind speed,air temperature,atmospheric pressure,relative humidity,etc.will all affect the wind power output power.Establish a prediction model based on influencing factors;due to the periodicity,nonlinearity,and non-stationarity of the output power time series,a single prediction model can make a good prediction of nonlinearity,but it is difficult to deal with non-stationarity,so a time-based Sequence short-term output power prediction method.In the establishment of prediction models based on influencing factors,an adaptive mutation particle based on mutual information(Mutual Information,MI)is proposed for the problem of low prediction accuracy and many input variables in wind power forecasting by BP(Back Propagation)neural network prediction model Group-optimized BP short-term wind power forecasting model.First,the mutual information is used to filter out the factors that have a greater correlation with the output power in the original data and reduce the redundant information contained in it.Then,the particle swarm algorithm that introduces the idea of adaptive inertia weight and mutation factor optimizes the prediction model Obtain short-term output power prediction value.In the establishment of a prediction model based on a time series,in view of the nonlinearity and non-stationarity of the output power of a wind power plant,a processing method combining empirical mode decomposition and wavelet threshold on the power time series is proposed to utilize the complex power time series EMD(Empirical Mode Decomposition)is decomposed into multiple modal(IMF)components,and then the high-frequency IMF component is subjected to wavelet threshold noise reduction,and then the low-frequency IMF component and the high-frequency IMF component are established based on least squares support vector machine and ARMA,respectively.(Auto-Regressive and Moving Average Model)The prediction model of the algorithm uses particle swarm optimization to optimize the parameters of each component prediction model,predict the power time series,and finally obtain the final prediction result by superposition.Finally,the meteorological data and historical output power data of a wind farm for 30 days are used as research samples to predict the short-term output power based on the two prediction models established above.At the same time,the mean absolute error(MAE),root mean square error(RMSE)and mean absolute percentage error(MAPE)are compared with other unimproved prediction models.It can be seen from the comparison that the method used in this paper predicts the output power more accurately.