Forecasting Models And Applications Of Operation Characteristic Parameters For Grid-connected Wind Turbine Generators

With the gradual increase in the installed capacity of wind turbines in China, windpower in the grid in the proportion of increased every year. Due to the uncertainty of themagnitude and direction of natural wind speed, grid-connected wind turbine oftendynamically switch between different operating conditions, the wind turbinecomponents, the stochastic nature of the increasingly obvious, plus in the harshenvironment under the conditions of the various components with the changes in theoperating environment and working hours due to the load, wear, fatigue strength andoperating performance will gradually decline. In order to reduce because of theirfailures, improve the reliability of grid-connected wind turbine is running, it isnecessary to predict and judge the characteristic parameters that reflect the operatingstatus of the wind turbine.Grid-connected wind turbine works in the analysis, control strategy and its impacton the basis, select the output active power, the generator speed gearbox input shafttemperature, generator winding temperature and inverter temperature as a reflection ofthe wind power the characteristic parameters of the unit operation status; pseudo-datato determine the characteristic parameters and Lagrange interpolation errors andomissions in the data amended and supplemented. This method is applied to the seamounted wind turbine wind speed, active power characteristic parameters of the data onthe9th amendment, the example shows that the Lagrange interpolation method isaccurate, simple and effective.BP neural network model parameters were established for each characteristic of thewind turbine, based on statistical data, the contact between the neural network layersand enumeration heuristics to determine the network layers and layers of nerve cell, andactive power and perfect correlation between the characteristic parameters of each state,the prediction model. Matlab platform programming. The model is applied to theprediction of the wind turbine characteristic parameters, numerical results show that theBP model has a simple, easy computation, parallel and strong and so.Were analyzed with BP neural network, based on historical experience data andenumeration heuristics were established for each characteristic parameters of the RBFneural network model, and K-means clustering method and the least squares method todetermine the radial basis function centers and connection weights, thus achieving a prediction of the characteristic parameters of the wind turbine. The model is comparedwith the BP neural network, numerical results show that the RBF model has a fastlearning speed, high precision, good convergence.Finally, the paper summarizes the advantages and disadvantages of variousprediction models and predict the results of analysis, further areas for improvement, andlooking forward to more wind turbine characteristics parameters of the forecastingmethods and subsequent related research.