Design And Implementation Of Power Demand Prediction And Load Characteristic Analysis System

The rapid development of China Southern Power Grid (CSPG) has made it become one of the world's most complex power systems. Various types of power sources within the system have been completed and the grid structure is very complex, but the internal resources and economic development are very uneven, therefore need to optimize resource allocation through a wide range of power transmission, which is bringing great difficulties to power grid planning and operation. Therefore, CSPG plans to develop one planning and operation database and some relevant advanced applications. Power demand prediction and load characteristics analysis are the essential work of power grid planning and operation; therefore they are the important parts of advanced database applications.The topic in this thesis is from the "Southern Power Grid Planning Operation Computation Analysis Database Project-Power Demand Prediction and Load Characteristics Analysis" research project of CSPG. The main tasks of power demand prediction and load characteristics analysis are to analyze the load characteristics and forecast electricity demand. The thesis studies the methods of power demand forecasting and load characteristics analysis to design and develop an advanced application accordingly.Firstly, a brief introduction of the basic concepts of electricity demand forecasting is presented and the commonly used algorithms for medium and long-term load forecasting are studied. Then the significance of important parameters, advantages and disadvantages of the prediction algorithm are analyzed. At last, the combination prediction methods are described, including model selection methods of combination forecasting algorithms and weight allocation methods.Secondly, load characteristics analysis theories are studied and Load characteristics indicators are introduced, including classifications, contents and significance. Then, the relevant factors of loads characteristics, including economic output, population, temperature weather, special holidays and large customers are studied. Lastly, the data mining methods, clustering analysis and gray correlation analysis, which are commonly used for load characteristics analysis, are described.Thirdly, an annual load curve forecasting method based on electricity error correction is presented. In the method, the year electricity prediction with relatively high prediction accuracy is applied to correct the annual load curve. The method is applied to compute one annual load curve of a certain area and the result is compared with the practical data to demonstrate the effectiveness and practicability. According to the requirements of grid work, one loads characteristics analysis method applying numerical rage as statistical conditions and one specific time load characteristics analysis method are designed, which makes up the disadvantage of the traditional load forecasting systems that load characteristics functions are weak.Finally, according to the research of this thesis, the request of "Southern Power Grid Planning Operation Computation Analysis Database Project-Electricity Demand Forecasting and Load Characteristics Analysis" research project and the standard data mining process CRISP-DM model, the main framework of the system, system database, and functional modules are designed and developed. The system, as a senior database application, providing application supporting for the work of network planning has achieved its desired results, which is important for the efficient operation and the long-term development of the power companies.