Forecast Of Wind Farms Short-term Output Based On Improved SVM

As wind energy is fluctuant, intermittent, low energy density, uncontrollable and other characteristics, wind farm power is also fluctuating and intermittent. The forecast of wind farm power output contributes to power grid scheduling department which timely develops logical daily system analysis according to daily load curve and can accurately adjust scheduling programs to meet the next day’s electricity demand. While the output of conventional units is optimized, the purpose of reducing the spinning reserve and the operating costs are achieved. The reliable, economical, and quality operation of power system is guaranteed.In recent years, there are many scholars using least squares support vector machine method to predict wind speed. The research purpose of this paper is to do further researches on this basis. Predict the output of wind farms by using the method of least squares support vector machine. This paper began from the basic principles of statistical learning theory, outlining the concept of linear and non-linear support vector machines, and analyzes the basic principle, learning algorithm, and model parameters selection of support vector machine regression problems.Different from standard support vector machines, LS-SVM converted the objective optimization function into a quadratic optimization problem, namely a quadratic loss function instead of the insensitive loss function in standard support vector machines. To study the forecast effect of LS-SVM for wind farm output power, this paper modeled on MATLAB software platform and made simulation experiments based on the wind farm historical data have been collected. This paper predicted the wind speed of the wind farm and compared the results with the standard SVM method prediction results.Identify the relationship between wind speed and wind farm output by using LS-SVM method which is based on the wind speed forecasting. Then the forecast of wind farm output is completed. The predicted effects of LS-SVM method and standard SVM method for short-term output of wind farms are analyzed and compared. It is found that LS-SVM method simplifies the calculation, improves the convergency rate and reduces the prediction error.Finally, the results of this paper are summarized. And the prospect for the future development of wind farm output forecast direction is made.