基于自适应模糊神经元网络的电力短期负荷预测

Electric Short Term Load Forecasting Based on Adaptive Artificial Neural Networks

利用模糊神经元网络（ＦＮＮ）进行电力短期负荷预测．给出了模糊神经元网络结构和部分输入变量的模糊化．ＦＮＮ采用ＬＭＳ（Ｌｅａｓｔ－Ｍｅａｎ－Ｓｑｕａｒｅ）算法，并用历史负荷数据进行训练．一经训练，网络就能应用于在线负荷预测．在预测过程中，权值按最近的负荷行为自适应调整．测试结果表明，该方法具有较好的精度和较快的速度．

A fuzzy neural network(FNN) for electric shotr term load forecasting is presented in this paper. A FNN structure is proposed and fuzzification of imput variables are given. The FNN is trained by historical load data with the LMS algorithm. Once trained, the FNN can be used to forecast future loads on line. An adaptive weight update strategy based on the most recent performance during the forecasting phase is developed. Test results show that the FNN can forecasting future loads with a higher accuracy and a faster speed.