论文部分内容阅读
灰色GM(1,1)预测模型,要求样本数据少,具有原理简单、运算方便、短期预测精度高、可检验等优点,在负荷预测中得到了广泛应用,但是也有其局限性。当数据灰度越大,预测精度越差,且不太适合经济长期后推若干年的预测,在一定程度上是由模型中的参数α造成的,为此引入向量θ,建立残差GM(1,1,θ)模型,利用蚁群优化算法对其进行求解,同时应用神经网络对其预测残差进行优化。实证分析表明,与传统的预测方法相比,大大提高了预测精度,该方法具有一定的实用价值。
Gray GM (1,1) prediction model requires less sample data, has the advantages of simple principle, convenient operation, high short-term prediction accuracy and testability. It has been widely used in load forecasting, but it also has its limitations. When the gray value of the data is larger, the prediction accuracy is worse, and it is not very suitable for prediction of several years of economic long-term postponement. To a certain extent, this is caused by the parameter α in the model, for which a vector θ is introduced and a residual GM 1,1, θ) model, which is solved by the ant colony optimization algorithm and the prediction residuals are optimized by using neural network. The empirical analysis shows that compared with the traditional prediction methods, the prediction accuracy is greatly improved. This method has some practical value.