基于模糊聚类的神经网络短期负荷预测方法

针对电力负荷的特点，综合考虑天气、日类型、历史负荷等对未来负荷变化的影响，提出了一种新的短期负荷预测方法。通过模糊聚类选取学习样本，采用反向传播算法，对24点每点建立一个预测模型。该方法充分发挥了神经网络和模糊理论处理非线性问题的能力，提高了学习效能，在负荷平稳的季节和负荷波动较大的季节都具有较好的预测精度。

SHORT-TERM LOAD FORECASTING USING A NEURAL NETWORK BASED ON FUZZY CLUSTERING

According to the features of power load and considering the combined influence of weather, day type and historical load data, a new short-term load forecasting method is put forward in which the learning samples are selected by the fuzzy clustering and using the BP algorithm, a forecastingmodel for each point in 24 points are established. This method gives sufficient play to the ability of processing non-linear problems by neural network and fuzzy theory, and the learning efficiency are improved. Using the presented method the better forecasted accuracy can be achieved not only for the seasons when the loads vary stably and slowly, but also for the seasons when the loads fluctuate drastically.