基于GM-GRNN的电力系统长期负荷预测

Long term load forecasting based on GM-GRNN in power system

由于长期负荷历史数据比较少,因此预测难度较大。在分析了灰色预测和神经网络预测的优缺点的基础上,提出了一种新型的预测方法——GM-GRNN预测方法,此方法就是将灰色预测方法和人工神经网络中的广义神经网络相结合的预测方法,新方法发挥了灰色预测方法中的“累加生成”的优点,能够削弱原始数据中随机性并增加规律性,同时避免了灰色预测方法及其预测模型存在的理论误差。最后采用我国某省年用电量的预测的算例表明该方法的预测精度优于单一的灰色预测和单一的神经网络预测方法,为电力系统长期负荷预测提供了一种有用的方法。

Because of lack of history load data, it is more difficult to predict long time load. The paper analyzes the merits as well as defects of grey prediction method and artifical neural network （ANN） method, and proposes a novel forecasting method named grey neural network. The new method utilizes the accumulation generation operation of grey prediction to transform original data and produce accumulated data. The data possesses better regularity which makes it easier to model and train the ANN and avoid the theoretical error of grey prediction method. Case study shows that this method is more accurate and faster than single grey prediction and single neural network method. It is a useful method for long term load forecasting.