地下水位粒子群优化神经网络组合预测模型

Combination forecasting model based on neural network optimized by particle swarm optimization for groundwater level

【目的】将粒子群优化神经网络组合预测方法引入地下水位预测中,以提高地下水位预测的精度。【方法】以回归分析法、指数平滑法、灰色GM(1,1)模型的地下水位预测结果及预测结果平均值作为网络的输入,以实际地下水位值作为输出,对3个单一模型进行非线性组合,建立地下水位的粒子群优化神经网络组合预测模型,应用实例对模型的预测结果进行了验证,并与3个单一模型及等权平均组合模型的预测结果进行比较。【结果】实例运用结果表明,粒子群优化神经网络组合预测模型的均方误差为0.740 9,平均绝对误差为0.657 6,均小于单一模型及等权平均组合模型的相应值。【结论】粒子群优化神经网络组合预测方法适用于地下水位的预测。

【Objective】The study was to improve the accuracy of groundwater level forecasting by applying neural network optimized by particle swarm optimization.【Method】In this method,the groundwater level predictions and the average forecast results of regression analysis,exponential smoothing and grey model were used as the network inputs,and the actual water level value as the outputs.Then the nonlinear combination prediction model of water level was established by combining three single models.Application instance on the prediction of the model results were verified with three single models and the equal weighted average model predictions were compared.The predicted results of the model was verified and compared with the results of the 3 single model and weighted average model.【Result】The example shows that the mean square error,average absolute error are 0.740 9 and 0.657 6 of the neural network optimized by particle swarm optimization combination forecasting method,less than the corresponding prediction error of a single model and the equal weight method.【Conclusion】The combination forecasting method of neural network optimized by particle swarm optimization is suitable for the prediction of groundwater level.