Application On Wind Speed Forecasting Of Wind Farm Based On RBF Neural Network

Nowadays, the technology of developing renewable energy has become one of the focuses around the world because of the gradual depletion of fossil fuels which bring about serious pollution. So wind energy, as a clean and renewable energy, gets worldwide attention and the majority of countries are trying to harness the wind power. However, the randomness and uncontrollability of wind energy may have an impact on the grid and bring about a series of security problems. To solve the problems of randomness and uncontrollability of natural wind, this paper forecasts the wind speed of the wind farm to improve the accuracy of the wind speed forecasting, to reduce the impact of wind power on the grid, to increase the security of the power system and to cut down operating costs of the power system.This paper studies the development and effects of wind speed forecasting, and also introduces the relevant knowledge. At the same time, it does not only present the realistic significance of wind speed forecasting, but analyses the current research status of wind speed forecasting on the farm as well. In the end, this paper establishes the wind speed forecasting model which takes the RBF neural network as a core and combines the genetic algorithm and gradient descent method to train the RBF neural network. This paper contains following parts:① This paper introduces the concept of wind speed and analyzes its distribution characteristics and regular patterns in detail. Then it introduces the methods of collecting and preprocessing wind speed related data, and makes a proposal to use linear interpolation method to amend missing test data. Finally, it makes a correlation analysis of wind speed and the relevant meteorological factors.② It gives detailed introduction of basic concepts of the RBF neural network and genetic algorithm. In order to improve the forecasting accuracy, this paper puts forward a forecasting model which combines the genetic algorithm and gradient descent method to train the RBF neural network. In the end, it makes simulated tests to two functions, and demonstrates the ability of global optimization of the genetic algorithm and the approximation ability of the nonlinear curve in the forecasting model.③ This paper applies the historical data of a wind farm to the forecasting model to establish the RBF neural network forecasting model, in which input data are temperature, humidity, pressure, wind direction cosine and wind direction sine, and output data is wind speed. The test result proves to be satisfying.④ It uses the optimized RBF neural network to design a simple simulation platform and visualizes forecasting data and figure to achieves human-machine interface and observe with ease.At the end of this paper, it summarizes the research contents. Meanwhile, it also recommends that what should be enhanced and improved in the subsequent study.