Research On Load Forecasting Model Based On Decision Tree

The short-term load forecasting is an important routine for power dispatch department, and one important part of Energy Management System. It is the premise and basic of economic and reliable operation of power system. Its precision directly influences power system's profit and quality. To study the load characteristic, the analysis theory and forecasting arithmetic and develop an efficient and practicable load forecasting system is becoming a very important issue currently. The paper introduces the data processing, forecasting arithmetic and the design and development of the load forecasting system.In order to accurately forecast the short-term loads of power system, the paper presents a new load forecasting method based on optimized decision tree, which efficiently takes the non-load factor'influences into account. The paper is constructs the load forecasting model at the basis of real course, finally obtains accurate forecasting results.In data processing, the paper firstly picks up the character curve which is embodiment of natural load curve by method of k-means, then checks and modifies the bad-data according as the character curve. The results are well. Then rough set is used to optimize the testing attributes of decision tree by reducing the non-load factors. In order eliminate limitation of the ID3 algorithm, an optimized algorithm MBID3 is presented to select the testing attributes. Comparing with the decision tree built by ID3 and MBID3 with the same example. In conclusion, the MBID3 is better than ID3.The paper according to the object-oriented method and using java to actualize k-means algorithm and decision tree algorithm, and set up an integrated system of decision tree. Good test results using actual data demonstrate that this algorithm can improve the accuracy of short-term load forecasting effectively, and it has superiority and practicability.