证据理论融合蚁群神经网络的短期负荷组合预测

为提高短期负荷预测的精度，引入了证据理论融合蚁群神经网络的组合预测方法，根据重庆市负荷的实际数据，采用蚁群神经网络作为单一模型对其进行初步预测，由BP神经网络对预测误差及主要外界影响因素进行分析建模，获得了每个模型的可信度，并用证据理论对可信度进行合成得到组合权值，进而实现对短期电力负荷的组合预测。结果表明，该方法拟合误差小、预测精度高，具有一定的应用价值。

Short-term Load Combination Forecasting Based on Combination of Evidential Theory with Ant Colony Algorithm-Neural Network

In order to improve the accuracy of short-term load forecasting, combination prediction method is proposed by combining evidence theory with ant colony algorithm-neural network. According to the actual load data of Chongqing City, ant colony algorithm-neural network as single mode1 is used to its initial forecast. Then the BP neural networks is selected to get the credibility of each model by forecasting errors and the main environmental influencing factors. And the evidence theory was applied to obtain the combination weight. So, short-term load forecast was fulfilled. Examples show that the fitting error of the method is small with high prediction accuracy and it has a certain application value.