以神经网络与模糊逻辑互补的电力系统短期负荷预测模型及方法

Short-Term Load Forecasting Model and Method for Power System Based on Complementation of Neural Network and Fuzzy Logic

根据电力系统短期负荷预测的特点,采用神经网络与模糊逻辑互补的方法建立了负荷预测模型。通过粗糙集理论中的信息熵概念对神经网络的输入参数进行了筛选,以与待预测量相关性大的参数作为输入,不仅减少了神经网络的工作量,缩短了计算时间,而且提高了预测的准确性;在神经网络中,通过引进动量系数和遗忘系数优化网络,提高了ANN的收敛速度;在模糊逻辑中,充分利用了人们对负荷变化取得的主观经验,引进不平均隶属函数,来反映负荷对温度的敏感性。

According to the characteristics of electric short-term load forecasting, a complementation method based on artificial neural network (ANN) and fuzzy logic is proposed to establish a load prediction model, which separates the forecasting job into two parts: one is basic load component, and the other is the component under temperature and holiday situation. The basic load component is forecasted by ANN without considering the effect of temperature and holiday, which reduces the work of ANN and simplifies its structures. The revision of basic load is finished by Fuzzy logic, only considering the influence of temperature and holidays. It makes full use of expert experience, and the final load forecasting result is obtained. By using Rough Set Theory, knowledge entropy is introduced to choose ANNs input parameters. Parameters with a high correlation are used for input ,which reduce the work and calculation time of ANN and improve the accuracy of prediction. Convergence speed of ANN is improved by using momentum factors and oblivion factors. While in the fuzzy logic aspect, an uneven membership function is used to describe loads sensitivity to temperature on the base of subjective experience of load variation.