一种基于ARMA-BP组合模型的电压偏差预测方法

国内许多变电站建立了电能质量的预测与预警机制，以应对日益严重的电能质量问题，其中电压偏差最为严重。针对预测模块中电压偏差预测算法的缺失，结合配电网的运行状态，提出了一种基于ARMA-BP组合模型的电压偏差预测方法。针对单一时间序列方法的不足，将时间序列和人工神经网络的算法结合起来。通过分析上海某变电站的电压偏差数据特征，首先采用时间序列的方法建立ARMA模型。然后采用BP人工神经网络的方法对ARMA模型预测值与原始数据之间的残差值进行拟合预测，最终得到2种模型预测所得累加值的结果。研究结果表明了所提方法的有效性。

A Voltage Deviation Prediction Method Based on ARMA-BP Combined Model

Prediction and early warning mechanisms have been established for substations in China to cope with the increasingly serious power quality problems， of which voltage deviation is the most serious problem. Aiming at the lack of voltage deviation prediction algorithm in prediction module， combined with the operating status of distribution network， the paper proposes a voltage deviation prediction method based on ARMA-BP combined model. For the shortcomings of single time series method， time series and artificial neural network algorithm is combined. By analyzing the characteristics of voltage deviation data from a substation in Shanghai， ARMA model is first established by the time series method. Then， the residual value between the predicted value of the ARMA model and the original data is fitted and predicted by the BP artificial neural network. Finally the accumulated value predicted by the two models is obtained. The results show the effectiveness of the method.