Research On Load Dynamic Characteristic Information Acquisition For Power System Modeling

Exact acquisition of load dynamic characteristic data is a problem to be settled urgently in load modeling for electric power system, since the precision of the load modeling data affects the veracity of the load model directly. In previous studies, the researches have to build model based on experience or history model because of lack of load dynamic characteristic data, so the model can’t describe the dynamic process of power grid correctly. In recent years, with people’s emphasis on load modeling and the increase of the load characteristic recording equipment, a huge of load dynamic characteristics data are accumulated, in face of vast amounts of load characteristic data, how to obtain accurate load dynamic characteristics data which can be used in load modeling has become a problem to be solved. In this paper, further research on exact acquisition of load dynamic characteristic data, main works of this thesis are listed as below.(1) Reliable efficient data storage is the premise and guarantee of data acquisition, the load dynamic characteristics data distributed in different geographical location, the data quantity is huge and the data has characteristics of distributed storage and autonomous, the scalability and reliability of conventional disk array storage and management methods is poor, and the data processing ability is insufficient, in order to solve current problems existing in the data processing of load dynamic characteristics data, a data storage and processing platform based on Hadoop cloud computing is designed and implemented, data are unified management under the same NameNode control. The platform provide technical support for distributed heterogeneous load dynamic characteristic data parallel processing analysis, it is easy to grasp the dynamic characteristics of load on high level.(2) Further research on processing of load dynamic characteristic data. Accurate data preprocessing is crucial for load modeling, this paper summarizes preprocessing method of the wide-area measurement system, fault wave record system and power quality monitoring system, and further research on bad data identification and correction and de-noising problem. In view of load dynamic characteristics data contains bad data, using longitudinal load dynamic characteristic data similarity on incidental in bad data identification, and using the parabola interpolation method for bad data modification; in view of load dynamic characteristic data consists noise interference by using wavelet analysis to remove noise. Verified the effectiveness and accuracy of the proposed method respectively. On this basis, in order to improve the efficiency of data access, set data processing platform based on Hadoop, designs the structure of platform and analyzing the realization of each module in detail. The platform provides a basic platform and foundation for load dynamic characteristics data processing.(3) Researched on dynamic characteristic data acquisition. Analysis of the traditional data analysis method of the data acquisition, and propose a method of acquisition of load dynamic characteristics data based on extreme learning machine. The simulations of the measured data of the power system verify the validity and rationality of extreme learning machine. By Comparing with the traditional classification method verify that the extreme learning machine has advantage of classification speed and classification accuracy, and the accuracy and speed of data obtained are improved. On this basis, complete the distributed parallel processing of data based on Hadoop platform, the timeliness of contrast between Hadoop platform and the single machine, prove the efficiency of Hadoop platform in dealing with data. Hadoop platform not only can solve the problem of mass load dynamic characteristics data distributed storage, and also solves the distributed parallel processing of the data, and it provides a new theory and effective methods and techniques for obtain load dynamic characteristic data which can reflect the real load characteristics.