Wind Power System Modeling Based On Artificial Neural Network And Its Application

With wind power occupying a considerable proportion, it is directly related to the stability of the entire grid. As distributed energy, it's modeling lack of systematic and universality research. Especially, wind power system structure is becoming more complex, nonlinear, and the characteristics of variability and uncertainty are more obvious, so, it is difficult to establish an appropriate mathematical model to articulate the relationship between variables. However, Neural Network based on non-mathematical modeling provides a new way of thinking to address the lack of mathematical modeling methods, and then, it has opened up an effective way for wind power system model.In order to solve the existing problem that the nonlinear of wind system and complicated factors, this paper using non-linear approximation properties of artificial neural network to build wind farms of static characteristic model. Firstly, the relationship with the parameters field data was analysised, using principal component analysis method to analyze field data, and data were normalized to fit the wind farm static characteristics model by artificial neural networkm, mainly uses the more classic model of BP network, but there are some unavoidable because of its problems, so the further use of RBF network to improve the static model.In order to better describe wind power system dynamic performance, and solve the problem that the most wind system load model separate active power, reactive power. Based on static model, using the neural networks for the stack weights but not the simple sum of mathematical sense, and using artificial neural network structure which can have multiple output characteristic, wind system can be regarded as load model, which its active power, reactive power these two output unifies in together. This is reflected that the power system load is mutual coupling active and reactive power, and through the data sample pretreatment, training and testing, enabling the network model to fit accuracy requirement, and can provide more accurate wind model, especially gives a good description dynamic behavior of the system.In order to provide a convenient use of user-friendly way, the article studies from the model application perspective using Visua1C++ and MFC to create an application that generates the window, combined with neural network structure characteristics, and packaged the weights, the threshold value to generate graphical user interface, and the formation of intermediate file to storage defined interface information. Graphical user interface program read and display interface information and displayed. The end, the user can interact with the application process the message by implementing the program which shows a graphical user interface, the drawing and provides a convenient and friendly way.