Research On The Wind Speed And Wind Power Forecasting Bansed On The Historical Meteorological Data

With the mass consumption of fossil fuels and the increasing environmental pressure,wind power has been attracting more and more attention from the countries all over the word as a renewable and clean energy.However,wind power has great intermittent and randomness and volatility as easily affect by the geography,meteorological and other environmental factors.It does harm to the power generation planning and economic operation of the power system.And it hinders the further promotion and application of the wind power.Therefore,it is based accurate wind power forecasting for the large-scale wind power grid operation.And accurate wind speed forecasting is the key factor in the wind power forecasting.It also has the important practical significance to improve the wind power forecasting precision.It is vulnerable for the wind speed time series affected by the effects of meteorological factors such as temperature,pressure,humidity and so on.Wind speed is difficult to forecast accurately as it has very high input dimensions and the strong nonlinear.Extreme learning machine(ELM)is an optimization model with the simple structure.It has many advantages as high operation efficiency,strong generalization ability and it is even can flexible to choose the hidden layers numbers and activation function types according to the issues.So it is suitable for the wind speed forecasting with the complex nonlinear.This article mainly aims at the study of the short-term wind speed and wind power forecasting and its relevant content is as follows.It is difficult to gain the precision forecasting results with numerous wind speed attributes.It is based on the mutual information to analysis the relationship between the attribute sequences and the wind speed and the power sequence.And it is used the maximum redundancy minimum redundancy(mRMR)attribute selection to reduce the dimension of the input.Then the Pearson correlation coefficient(PCC)is adopted to weight the input attributes in order to highlight the degree of the high degree attribute sequences' importance to improve the prediction precision.Finally,the ELM and its optimization model are used in the prediction research.In order to reduce the instability of wind speed time series and its nonlinear influence on the power system,the signal decomposition methods can be used to get the relatively stable subsequence for the further improvement of the the wind speed forecasting accuracy.This article uses the variational mode decomposition(VMD)to deal with the original time series.Then it can get a series of relative stable sequences with certain periodicity and regularity.The partial autocorrelation function(PACF)is used to select the higher correlation degree of the set as the inputs of the network model for each subsequence.Then the weighted regularized extremelearning machine(WRELM)is selected to build the multi-step rolling forecasting modelto carry out the short-term wind speed forecasting with its generalization ability.Finally,this paper uses software MATLAB 8.5(2015a)to simulate the methods above with the measured wind speed and wind power data.And it verifies the validity and practicability of the new methods.