基于卡尔曼滤波的短期负荷多步预测修正模型研究

提出了一种短期负荷多步预测的修正方法。首先采用BP神经网络法建立短期负荷的分时多步预测模型，对于每一个初始预测值，采用卡尔曼滤波模型进行修正，以减少模型的累积误差，提高多步预测的效果。算例结果证明了所提方法不仅能够提高单步预测的预测效果，而且能够有效降低多步预测的误差，对于实现连续日短期负荷预测具有现实意义。

Study on Modified Model for Multi-step Forecasting of Short-term load Based on Kalman Filter

This paper proposes a modified method for multi-step forecasting of short-term load. Firstly, the BP neural network method is adopted to establish time-sharing and multi-step forecasting model of short-term load; then Kalman filter model is utilized to modify each initial forecast value to reduce the cumulative error of the model and improve multi-step forecasting. The calculation example result demonstrates that the pro-posed method can not only improve forecasting of single-step forecasting but effectively reduce multi-step forecasting errors;it is of operation significance for consecutive daily short-term load forecasting.