基于D-S证据理论的母线负荷预测

Bus Load Forecasting Based on D-S Evidence Theory

基于D-S证据理论,提出一种母线负荷预测新方法。该方法对BP神经网络、改进的BP神经网络和支持向量机(SVM)的母线负荷预测模型分别建立权重提取和权重融合模型,并运用D-S证据理论对3种预测模型的权重进行融合。通过对预测数据进行分析,提取证据理论样本,并将可信度函数的融合结果作为母线负荷预测模型的权重,最终得到待预测日的母线负荷预测结果。仿真结果表明,与单一的母线负荷预测模型相比,经D-S证据理论融合的母线负荷预测模型更有效,也具有更高的预测精度。

In this paper, based on Dempster-Shafer （D-S） evidence theory, a new bus load forecasting approach is presented. The proposed method will firstly establish the weight extraction model and weight fusion model for BP neural networks, generalized BP neural networks and SVM bus load forecasting models, respectively. Then, by using D-S evidence theory, the weights of three forecasting models are fused. After the extraction of fusion samples of evidence theory by analyzing forecasting data, and multi-fusion result of belief function is taken as the weight of bus load forecasting model, by which the bus load in the future is forecasted. Finally, simulation results provided later show that the proposed method in this paper are more effective and has a higher forecasting accuracy than that by using only one model.