Research On Wind Power Prediction Method

With the development of the national economy,the development of wind power in China has been rapidly promoted.As a green renewable energy,wind power has now received extensive attention and policy incentives,and it has been incorporated into the national long-term energy development strategy.due to changes in wind speed and direction,wind turbines generate large randomness,fluctuations,and instability,which adds many difficulties to the rapid,effective and safe dispatch of the power grid,and seriously affects the safe operation of the power system.Accurate short-term prediction of wind power can realize the reasonable distribution and dispatch of electricity by wind farms,effectively develop and utilize wind energy,and ensure the stable development of the power industry.Variable Mode Decomposition(Variational Mode Decomposition,VMD)is a new adaptive signal processing decomposition technology proposed in recent years,which can process non-linear and non-stationary signals.This paper uses variational mode decomposition to decompose the original wind power data signal,and combines the improved grey wolf optimization algorithm and the kernel extreme learning machine neural network to make predictions.In order to improve wind power forecasting accuracy,this paper uses actual wind farm data to conduct short-term wind power combination forecasting research.To study the advantages of the combined prediction model in the prediction of non-stationary data signals,the main research contents of this article are as follows:1.The original wind power data is pre-processed.Due to the missing or abnormal values of the original collected wind power data,according to the rationality test of the wind power data,the abnormal data is processed by the quartile method to make it meet the standard wind speed-power of wind turbines curve.Aiming at the volatility and instability of the power data,a signal decomposition algorithm is used to perform noise reduction processing on the processed wind power data to reduce the impact of data non-stationarity on the prediction model and improve the performance of the prediction model.In this paper,the Variational Mode Decomposition algorithm is used to decompose the time series data of wind power,and a limited number of relatively stable intrinsic modal components with different bandwidths are obtained to improve the data quality.This is also a prerequisite for high-precision modeling.2.Two prediction methods,BP(Back Propagation,BP)neural network and deepbelief network are used to predict and analyze wind power data.By analyzing the prediction effect maps and prediction evaluation indicators of each algorithm,it is found that the prediction curve is close to the trend of real value fluctuations.The prediction effect of the deep belief network is better than the BP neural network,and the prediction performance is optimal,but the prediction accuracy needs to be improved.3.Aiming at the low prediction accuracy of a single prediction model,a combined prediction model based on a combination of a kernel extreme learning machine optimized by an improved grey wolf algorithm and variational mode decomposition is used,and the power data of the wind turbine is predicted.The grey wolf optimization algorithm can be used to optimize the kernel parameters and penalty coefficient in kernel extreme learning machine,effectively improve the model prediction accuracy.The experimental results show that compared with the neural network model without the optimization algorithm,the combined prediction model has improved prediction accuracy.