针对电力负荷受天气和日期影响特点，提出一种基于动态模糊神经网络的短期电力负荷预测的新方法。该算法最大的特点是模糊规则是动态变化的，通过系统误差、可容纳边界来判断系统是否需要新增一条模糊规则，使用误差下降率（ERR ）修剪算法剔除对整个网络影响较小的模糊规则。该算法还使用了分级学习法让网络的学习速度大大提高。在分析了EUNITE网络提供的负荷数据基础上来进行仿真，该仿真将温度、星期、月份、节假日因素作为网络的输入向量，取日负荷峰值作为网络的输出向量。仿真结果显示取得了较好的预测准确率。%According to the characteristics of the power load affected by the weather and the date ,this paper puts forward a new method for short-term load forecasting of power system based on dynamic fuzzy neural network. The main feature of this algorithm is that its fuzzy rules are dynamic. It can analyze the systematic error and the boundary to decide whether it needs to add a fuzzy rule, and then use the error reduction ratio (ERR) to remove fuzzy rules which have small impact on the whole network. The algorithm also uses a hierarchical learning which greatly improves the learning speed of the network. Based on analyzing the EUNITE provided by the network, the simulation takes temperature, week, month, holiday factors as input vectors of the network, and the daily load peak as the output vector of the network. Simulation results show that the algorithm has good forecast accuracy.
展开▼
