粗糙集信息熵与自适应神经网络模糊系统相结合的电力短期负荷预测模型及方法

电力系统短期负荷预测不仅要考虑负荷本身的历史时间序列,而且与气象因素密切相关,自适应神经网络模糊系统(Adaptive Neuro-Fuzzy Inference Sysrem,ANFIS)模型是一种有效的预测方法,而系统输入变量的合理性选择是影响预测效果的关键所在.作者通过粗糙集理论中的信息熵概念对解决这一问题进行了尝试,选取与待预测量相关性大的参数作为输入.所构造ANFIS系统是基于数据进行建模并进行参数辨识的,这样有效地避免了模糊推理系统(Fuzzy Inference System,FIS)中人为主观因素对预测的负面影响,客观地反映了相关变量与负荷值之间的复杂关系.用该方法与常用BP神经网络及常用FIS分别对重庆市某区进行了一周的日负荷预测,通过对实例的对比分析表明了该方法具有较好的收敛性和预测精度.

MODEL AND METHOD FOR POWER SYSTEM SHORT-TERM LOAD FORECASTING BASED ON INTEGRATION OF INFORMATION ENTROPY IN ROUGH SET THEORY WITH ADAPTIVE NEURAL FUZZY INFERENCE SYSTEM

In power load forecasting not only the time sequence of load itself should be considered, but also the influence of weather factors should be taken into account. Adaptive Neural Fuzzy Inference System (ANFIS) is an effective method in the modeling of mufti-factor load forecasting, however, in which the key step that influences the accuracy of forecasted results is the reasonable selection of the variables. To solve this problem, the conception of information entropy in rough set theory is used to select the variables which have close relativity with the output parameter as the input parameters. In the constituted ANFIS the modeling and system parameter identification are based on the actual data, so the negative interaction of man-made subjective factors in Fuzzy Inference System (FIS) is effectively avoided and the complex relations among relevant variables and load values can be objectively reflected. The daily load forecastings during a week for a certain district in Chongqing city are respectively carried out by the presented method, the BP neural network and the FIS in common use. Through the analysis of the obtained forecasting results it is shown that the presented method possesses better convergence and is more accurate.