Study On Short-term Wind Forecasting With Chaotic-support Vector Machine In PcDuino Platform

In recent years, China is facing with dual pressure of environmental pollution andenergy shortage, so green, low-carbon wind power technology is widely applied. Taking intoaccount the presence of wind power penetration power characteristics, we carry out short-termprediction of wind speed and calculate the wind power according to the wind speed, so we cantake steps in advance to ensure safe and stable operation of the power grid. This thsis takescare of the combination of chaotic phase space reconstruction and support vector regression toanalyze and forecast the wind data, realizes Chaotic-Support Vector Machine short-term windforecasting on PcDuino Platform.Firstly, the thesis introduces the basic concepts of wind and wind speed, and analyzes thevariation of wind speed and direction in the general sense, describes the count of wind energyand wind power generation, analyzes the characteristics of wind speed and direction collectedfrom the field, discusses the reasonable selection of the number of SVM training set.Secondly, the thesis introduces the concept of chaos theory, phase space reconstructionof chaotic time series and determines the chaotic characteristics of the chaotic time series.Then derives the process of support vector regression and introduces the prediction specificimplementation steps by the method of phase space reconstruction and support vectormachine. A wind forecasting system is designed with a user interface of Matlab GUI. Thissystem can quickly import, switch, display and processing of wind data, and can be a key toachieving the chaotic support vector machine, time series and chaotic RBF forecasting orother relational functions.It greatly improves the efficiency of wind prediction.Thirdly, it is necessary to decompose the wind data to X and Y directions to analyze andpredict. Time delay and embedding dimension are obtained using the mutual informationmethod and the CAO method. The Cross Validation-Grid Search method is used for SVMparameter optimization. Analyzes the chaotic characteristics of wind time series in different time intervals. Investigate and study the predictive model using the average relative error andthe mean absolute error.The predictive results are compared with the RBF neural network andthe time series prediction model. The comparison of results show that the the combination ofchaotic phase space reconstruction and support vector regression is a small error, highprecision forecasting method.At the last, to make Chaotic SVM into practice, embedded achieves wind forecasting byOctave software in pcDuino platform，describes the process of porting the code to predict thewind speed and the overall implementation process in pcDuino platform.