基于竞争分类的神经网络短期电力负荷预测

根据电力负荷的特点,在考虑天气、日类型、实际历史负荷等因素对预测负荷影响的基础上,提出了一种基于竞争分类的神经网络短期负荷预测方法.应用神经网络的竞争学习对相关数据进行分类,将历史数据分成若干类别从而找出与预测日同类型的预测类别.利用相应的BP算法对未来24小时负荷进行短期预测,该方法充分发挥了神经网络处理非线性问题的能力.结果表明,该方法取得了较满意的预测精度.

SHORT-TIME LOAD FORECASTING USING NEURAL NETWORK BASED ON COMPETITIVE LEARNING CLASSIFICATION

According to the features of power load and considering the influence of weather, day type and practical historical load on load forecasting, a short-time load forecasting method using neural network based on competitive learning classification is presented. Applying the competitive learning of neural network to the classification of related data, the historical data is classified, thus, the historical load data is divided into several load sorts and the load sort to be forecasted, which is as same as that of the forecasted day, can be found. The corresponding BP algorithm, which brings its ability of processing non-linear problem into full play, is used to forecast the short-term load in future 24 hours. The forecasting results show that the forecasting accuracy is satisfied.