Study On Data Mining Application In Short-term Load Forecasting

Power system short-term load forecast is an important work of electric power system dispatching operations. The level of forecast accuracy directly affects the operation of the power system security, the economy and quality of power supply. The short-term load forecast is embarks from the known social economy development and the electric power demand, through to the historical data research analysis, day or a week-long system load will carry on the estimate and the extrapolation to the future. Load factor affecting many of these factors on the impact of non-linear loads, complexity, and the characteristics of the lag. If no analysis of factors correct, to be precise about the load forecasting is a very difficult decision.To accurately predict the load, the need for an in-depth analysis of historical data, as data mining technologies in the mining vast amounts of information from the knowledge, and therefore the introduction of the load forecasting data mining theory, with a view to establishing accurate forecasting model. Adopting various kinds of data mining unit model, data mining engine guide on historical data warehouse, excavation to discover useful knowledge. These include knowledge of the impact of changes in load factors, factors change, load change the law, and forecast information on the importation of the most suitable training samples and forecast samples. In the prediction model and the choice model algorithm and give full consideration to seasonal, weather, temperature and holidays, and other factors, step by step means clustering algorithm using the above factors multi-level details of decomposition clustering and classification. Artificial neural network model will be introduced to the short-term load forecasts, historical data samples selected by BP network algorithm for training. On the impact of short-term electric load the important factors weighted by the actual load data to verify the dynamic adjustment of high prediction accuracy.Network scheduling department for the actual needs of the development of a scheduling automation system based on the short-term power load forecasting system. The system integrated into the regional power grid scheduling automation systems, with better real-time performance, economy and practicality, using object-oriented design methods and client/server development approach of the load forecasting system completed the structural design of framework, and the function and main features. And from the forecast, data management functions, as well as some auxiliary functions, such as point of the system to conduct a comprehensive briefing.