The Method Research Of Wind Turbines Operating Conditions Assessment Based On Big Data Analysis

With the rapid growing of wind power industry in the world, more and more wind power penetration into the power system, affecting both stability and dispatching of traditional power grid. The accurate assessment of the operating conditions of wind turbines, not only to their own operation, but also of great significance to the security and stability of power grid. At present, most wind turbines are equipped with the SCADA system, which can collect and record a full range of wind turbines operating conditions information, providing a reliable source of big data for the assessment of wind turbines operating conditions. In this case, how to effectively analyze monitoring data, and carrying out operating conditions assessment, is the focus of concern to "intelligent" wind turbines. Therefore, in order to study operating conditions assessment methods of wind turbines, this paper carries out the following researching works:(1)The operating conditions predicting model of wind turbines is proposed. From the perspective of error correction, this paper has established a combined predicting model based on Markov chain(MC) and Support vector regression(SVR), to predict very short-term wind power. Firstly, based on monitoring data of SCADA system, the SVR is used to model for predicting wind power, and calculate its predicting errors.Then, the conditions transition probability matrix is made based on MC to modify the wind power predicting results of SVR model. Finally, the confidence level of wind power correction results predicted by SVR-MC is given by method of fluctuation confidence interval.(2)The system operating conditions assessment model of wind turbines is proposed.Based on dynamic inferior degree, this paper presents a novel real-time system operating conditions assessment method. Firstly, the concept of dynamic inferior degree is introduced, which can dynamic describe the changing of wind turbines operating conditions by taking the variation of assessment indices into the calculation process of inferior degree. A Markov chain model is adapted to predict the variationof assessment indices, which needs only the current data in predicting process and can meet the requirements of real-time assessment. Then, inputting the dynamic inferior degree into normal cloud model, which overcomes the shortcoming of normal cloud model in reflecting the deviation of the membership degree. Finally, applying the normal cloud model into fuzzy synthetic evaluation method, a real-time operating conditions assessment method is given.(3)The wind turbines fault predicting model is proposed. The gearbox is one of the important components of wind turbines, which has highly costs of operation and maintenance, this paper presents a novel fault predicting model for gearbox, which has practical value. Firstly, based on SVR algorithm, the regression predicting model of gearbox operating temperature is established, to obtain the relative errors sequence of the predicting values, which describes the changing trend of gearbox. Then,inputting the relative errors sequence into improved backward normal cloud generator with the non-certainty degree, the digital features of normal and abnormal cloud model are obtained, meanwhile, the gearbox conditions cloud models are also given.Finally, operators can use the real-time monitoring data of SCADA system to calculate the closeness degrees of normal and abnormal cloud models, and take the principle of maximum closeness degree to complete fault prediction.Above three researching models, including the wind turbines operating conditions prediction, system operating conditions assessment and fault prediction respectively,are entirely based on real operating data of SCADA system, and the researching processes are also simple. What’s more, the researching results meet the needs of project, which has a good application value.