Study On Output Characteristics Of Wind Farms Considering Predictive Model

It is an effective way for the wind energy exploitation to integrate large scale wind power to the power grid. Since the fluctuating and almost uncontrollable characteristics of wind power, it is difficult to achieve high precision of wind power forecasting. Thus, large-scale wind power integrating to the grid will weaken the power generation control, which may threaten the security and economy of grid operation. In case it is difficult to accurately grasp the adverse affects of wind power fluctuations, the grid may overestimate the risk, thus restricting the scale of wind power to ensure the safe operation of power grid, which is to some extent hindered the development of wind power. With the construction of many 1000MW-class giant wind power bases in China, how to meet the needs of wind power development and ensure power grid security is an urgent task to be studied with both important theoretical and practical significance. It is necessary to get better knowledge of wind power prediction to mitigate the impact of wind power randomness on the grid, thus-the dispatch department can make accurate implement scheduling.First, there is a certain correlation in time and space of wind farm power fluctuations, which has the general characteristics of a random phenomenon. It is possible to describe the statistical regularities and trends from the mathematics. This article describes the wind power fluctuation characterization and evaluation methods, based on which, the statistical description of wind speed and power fluctuations of a Northeast wind power base was given within a certain period of time.Second, this paper proposed a method of wind power prediction based on correction factor after in-depth understanding of wind farm power fluctuation characteristics. Spatial and temporal distribution model of wind speed was established to obtain the correlation between the front and rear wind speed. Then, the wind power prediction accuracy was adjusted by the correction coefficient. The superiority of the proposed method was validated based on the actual wind speed data in a Northeast wind farm compared with BP neural network method.At last, the voltage distribution characteristics of wind farms was the key research of this paper based on the wind power predicted by corrected coefficient method. Reactive voltage distribution formula of doubly-fed wind turbines and wind farms was deduced in this paper, then detailed wind farms model for wind power prediction using the correction coefficient method was built in DIgSILENT to research the influence of wind power output fluctuations on wind farm voltage, considering different operating conditions and the access locations in an illustrative example. The analysis results of this paper can provide a reference for grid planning and dispatching.