Research And Implementation Of Decision Tree-based Short-term Load Forecasting System

Short-term load forecasting of power system is the base of power system security, affecting grid economic operation, especially for China's current shortage of electricity production in some areas and the low level of electric power dispatching. Power system generation side, DEH-side and the electricity market in the future all have high standards on it. For every 1% increase in short-term load forecasting, enormous economic benefits would be brought to the power industry, and would make a significant contribution national energy conservation emission reduction.The load itself is great uncertainty and impressionable to external non-load factors. Now, we still haven't got a load forecasting model that adapt to the specific situation in all regions. That is, a model which performance well in many applications, may also not suit to another special circumstance application. It is necessary to carefully study the data of the region that required predicting and find out a suitable model.In this paper, author study the true history data of the Hainan power grid load, choose the decision tree method to predict the 96-point daily load, tuning the ID3 algorithm and implement it. The specific works are as follows:1. Hainan's climate is relatively simple. There's little difference in temperature throughout the year, and only few factors affect the load. So decision tree is a good solution to predict 96 points load in Hainan.2. In the implement of ID3, a few adjustments have been made: 1, to the greatest extend, use static nodes instead of dynamic nodes, to simplify the generation of the tree, to reduce the burden of the server and shorten the load forecasting time; 2, also make corresponding adjustments in the selection of target properties, using the clustered max load instead of relative increment of load.3. Developing Hainan power grid load forecasting system and the decision tree prediction module. The system is based on B / S develop model. Struts2 and dwr framework have been used in this system to promote facilitate of the popularizing in different operating environments and applications. System is proved to have friendly interface, simple operation and strong portability.