Wind Power Prediction Based On Gray Scale Combined Algorithm

Traditional power generation mode shows its competitiveness with the uncertainty and volatility of the wind power, while the wind energy prediction and power storage prediction are two main effective solutions. Compared with the energy prediction, power prediction receives increasing attention with its easier industrialization and lower cost. In this article, the wind power generation are researched by a certain wind farm in China, together with the genetic algorithm, the least squares support vector machine, the grey algorithm and other algorithms.In the second part, the principle of the wind power generation is introduced in detail. The wind power is mainly determined by the power, direction, and some other factors, so the indirect prediction is adopted to predict wind farm power according to the known factors.The wind power prediction is mainly based on the meteorological data, and this article proposes the time series of power to predict the historical data. The first part presents the least squares support vector machine(LSSVM) predict on the ultra-short term wind power, while the second part improves accuracy and efficiency by the parameters prediction according to the genetic algorithm(GA) and the algorithm of gray method(GM). This processing gets a satisfied result and improves the 5%~8% compared with the traditional neural network optimization algorithm.According to the high uncertainty and volatility characteristics of the wind farms power, this article makes the short term wind prediction by an amount of unclear information on the gray system. So the results are subtracted from the tested signal to get the residual. The GM-LSSVM combination algorithm is that the LSSVM residual prediction is added to the GM part to get the final prediction results. The precision of the proposed method is improved when compared with the RBFNN and GA-LSSVM methods in simulation data.The combination algorithm with GA-LSSVM and GM-LSSVM is proposed to overcome the adaptability shortage under different wind fields and climatic conditions, which is more efficient, stable and accurate. As a result, the MAPE is lower than GA-LSSVM and GM-LSSVM by 0.45% and 0.2% in turn.The correlated gray algorithm effectively improves the fitness with the remaining accuracy in the experimental results, solves robustness in wind farm output, and improves the stability of power grid operation. The research results show that the proposed method is feasible in the wind farm output prediction.