Study On Wind Speed Soft Sensor For The Wind Power Generation System

With the development of wind power generation technology, how to reduce thecost of wind generation to a safe and efficient operation is the current focus.Measuring the effective wind speed accurately is the basis and key of operationcontrol. However, because the wind generation system exists in a three-dimensionaland time-varying wind farm, and that the distribution of wind speed is differenton the plane of blade rotation, the value of wind speed measured from anemometeris not accurate. This subject is to study a highly effective method for measuringwind speed, which can accurately predict the wind speed of wind power generation,and that improve the control capacity of wind turbine.Firstly, this paper introduces the background and current research situation ofwind power technology, analyzes and compares the methods of wind speedmeasurement in the wind power generation system, and that proposes a soft sensormethod of effective wind speed based on sopport vector machine.Secondly, according to its mathematical model and control strategy, simulationmodel of permanent magnet direct drive variable pitch wind power generationsystem is established, and that according to design requirements of soft sensor,dominant and secondary variables of the effective wind speed soft sensor model aredetermined. In order to obtain the sample data that is similar to the real situation,based on short-term wind speed characteristics, the method of wavelet transform isapplied for numerical simulation of turbulent wind speed spectrum to obtain thewind speed sequence in line with the Kaimal spectral characteristics. Aftercollecting and preprocessing the sample data from simulation model, wind speedsoft model based on the support vector machine is to be established.Finally, in order to improve the accuracy of effective wind speed soft sensormodel, the optimal method that combines genetic algorithm with simulatedannealing algorithm is proposed on account of the parameters of support vectormachine. With the simulation of grid search method, genetic algorithm and geneticalgorithm combined with simulated annealing algorithm, the result shows that theeffective wind speed soft sensor model based on support vector machine withparameter optimization owns more excellent capability of searching and the higheraccuracy of prediction.