不同峰值时段的主动配电网短时负荷可信度预测方法

Short-time Load Reliability Prediction Method for Active Distribution Networkat Different Peak Periods

无法确定预测模型权重会导致负荷可信度预测精准度较低,针对该问题,提出不同峰值时段的主动配电网短时负荷可信度预测方法。通过计算不同时间序列之间最优映射方式研究不同峰值时段的相似性,根据动态时间规划确定最优规整路径,并确定不同峰值时段,以此构建预测模型。确定拟合误差与发展相关度的相对权重,在组合预测中引入等维信息思想,使用灰色关联度法计算各单项模型的相对权重,并设计组合预测流程。在确定分配系数的基础上,将神经网络的输出直接送入馈线层,设计母线负荷预测流程。由算例分析结果可知,该方法预测最大误差为1.5 MW,预测精度较高。

If the weight of the prediction model can not be determined,the accuracy of load reliability prediction will be low.Aiming at this problem,a short-term load reliability prediction method for active distribution network in different peak periods is proposed.By calculating the optimal mapping between different time series,the similarity of different peak periods is studied.According to the dynamic time planning,the optimal regularized path is determined,and different peak periods are determined.Based on these,a prediction model is constructed,and the relative weights of fitting error and development correlation are determined.The idea of equal dimensional information is introduced into the combination prediction,and the single model is calculated by the grey correlation method.And the combination forecasting process is designed.On the basis of determining the distribution coefficient,the output of the neural network is directly sent to the feeder layer,and the bus load forecasting process is designed.The results show that the maximum error of this method is 1.5 MW,and the prediction accuracy is high.