基于模糊推理系统的多因素电力负荷预测

MULTI-FACTOR SHORT-TERM LOAD FORECASTING BASED ON FUZZY INFERENCE SYSTEM

在多因素负荷预测的建模中 ,模糊推理系统是一种较为有效的方法。输入变量选择和输入空间划分是模糊建模的基础 ,也是难点所在。数据挖掘中的分类思想有助于解决此问题。文中采用分类和回归树 ( CART)算法对解决这一问题进行了尝试 ,并构造 ANFIS网络进行参数辨识。建模过程几乎完全基于数据进行 ,不需要人工的过多干预 ,保证了模型能客观地反映相关变量与负荷值之间的复杂关系。用该方法与普通 BP算法分别对浙江省多地区进行了一个月的日负荷预测实例分析 ,该方法较好的收敛性和预测精度说明 ,CART算法与 ANFIS相结合 ,是基于数值的模糊建模的一种有效方法。

Fuzzy inference system (FIS) is an effective method for multi-factor load forecasting. Selection of input variables and space divisions is the premise and also the difficult problem for modeling the FIS. CART is introduced to solve the problem. And then an adaptive network is constructed to identify the parameters. The FIS is built completely based on the sample data, so the relationship between the input variables and the load value would be objectively reflected. The method of CART-ANFIS and BP are both used to forecast the next-day load in three areas of Zhejiang Province, China for a month. Better convergence and precision of CART-ANFIS are served for demonstrating that the modeling of FIS based on the sample data can be successfully accomplished by combining the CART and ANFIS.