Wind Power Distribution Pattern And Probabilistic Prediction Method

With the world-wide concentration on the problems, such as environmental protection, energy conservation and emission reduction, sustainable development, and so on, renewable power generation technologies capture more and more people's attention. From the technical and economic perspectives, wind power generation is one of the fastest growing and the most mature renewable energy technologies. In recent ten years, wind power cost has been greatly decreased, and wind power generation already has the potentialities of competition with traditional and conventional power generations, because of which many countries integrate wind power into the state development planning as a measure of energy structure improvement and ecological environment protection. As we all know, the randomness, volatility and uncontrollability of naturewind make the output power of wind turbines fluctuating, so when the wind power penetration reaches a certain proportion, the fluctuation will seriously influence the safe and stable operation of power grids. Therefore, in order to optimize the electric power dispatching, reduce the spinning reserve capacity, improve the wind power penetration limits, meet the power market transaction demands, arrange the maintenance and repair of wind turbines conveniently, the research on wind distribution characteristics is necessary, which is to strengthen the relevant application direction. Wind speed distribution characteristics of wind farms, obtained by the statistic of long term wind speed sample data, currently are applied to wind farm or electrical power planning. But when it comes to the situation, which needs to deal with shorter term, especially problems related to the time level of operation and control, the distribution characteristics mentioned above can't fully meet the new demands, on which the in-depth research from a new perspective is needed.Firstly, in order to cope with this situation, according to different application environments, pattern differences of wind speed distribution characteristics for wind farms were discussed. Analyzing main possible factors causing pattern differences and extracting features, from normal probabilistic density and conditional probabilistic density, such as characteristic wind speed, shape parameter, scale parameter, probabilistic deviation and so on, pattern differences given rise to different distribution performance were investigated. The wind speed distribution characteristics analysis results of a real wind farm validated the conspicuousness of pattern differences from different angles, which meant each application should adopt its related statistical law.Secondly, according to the different pattern features mentioned above, an appropriative wind power distribution pattern library was established to convenience the searching and matching operation on the historical patterns for users. The library included four modules: management, update, merging and user. Among them, the management module, measuring and matching patterns using the similarity function proposed in this paper, was responsible for the establishment and maintenance of the library. The update module was used to update the pattern library by adding the new wind data, produced with the operation of the farm, into the historical database, to enrich the library and accurately reflect the wind farm characteristics. Under the circumstance of pattern redundancy, the merging module was activated to streamline the library and maintain the practicability and efficiency of the library. The user module was designed to coordinate with the management module to process interactions of users and the library.Finally, a deterministic model based on the time series method for the actual wind farm data was established to predict the future wind speed data by one step. Then, the probabilistic prediction results were obtained by matching the current pattern with the wind farm pattern library and integrating with the wind generator power characteristic curve. The probabilistic prediction information can be used to meet the short-term wind power peak adaptation of the power system and the reasonable system spinning reserve arrangement demands.The main research is supported by the National Natural Science Foundation of China under Grant No. 50877014. The research conclusions are expected to make directive significance to the construction and operation of wind farms and the operation and control of the injected electric power system.