基于最大偏差相似性准则的BP神经网络短期电力负荷预测算法

针对企业电力负荷随机性强、稳定性低、预测精度不理想等问题，提出了一种基于最大偏差相似性准则的BP神经网络短期电力负荷预测算法。首先对最大偏差相似性准则算法进行修改，并提出使用预测日的负荷特征向量与最大偏差相似性准则算法聚类之后的类中心负荷特征的距离来确定预测日的相似日类别；然后将聚类后的相似日类别负荷数据作为BP网络的训练数据，输出预测日起始的连续三天96整点负荷值。实验表明，该方法提出的短期电力负荷预测方法在精度和网络训练时间上都有较大的提升，具有较高的有效性和实用性。

Short-term power load forecasting algorithm based on maximum deviation similarity criterion BP neural network

This paper proposed a short-term power load forecasting algorithm based on maximum deviation similarity criterion BP neural network to solve the problem of strong randomness, low stability and poor prediction accuracy of enterprise pawer load. This method first modified the maximum deviation similarity criterion algorithm, and proposed to use the load feature vector of the forecast day and the distance of the center load characteristic after clustering to determine the similar day class of the forecast day. Then, it used the similar daily class load data after clustering as the training data of BP network, and output the load of three consecutive days for 96 points. The experiment shows that the proposed short-term power load forecasting method has great improvement in precision and network training time, and has high effectiveness and practicability.