ARIMA-MSFD组合模型在甘肃省水力发电量预测中的应用

Application of ARIMA-MSFD combination model in forecast of hydropower generation in Gansu province

为了进一步提高电力负荷预测准确率,弥补单一模型自身局限性,组合模型被广泛应用。由于预测模型种.类较多,合理地选择适合的模型进行组合显得尤其重要。考虑到神经网络中可能包含多余的自变量和不合理的网络结构都会严重影响到整个网络的学习精度和求解速度,本文使用ARIMA时间序列模型与MSFD神经网络模型进行组合,修正预测结果。为了验证组合预测模型能够得到更高的预测精度,本文分别应用ARIMA时间序列单一模型、MSFD神经网络单一模型和组合预测模型ARIMA-MSFD模型对2008年到2018年的甘肃省水力月发电量进行学习,并以此预测2019年间12个月的发电量数据,通过对比三种模型预测结果的MAPE(平均绝对百分比误差)、MAE(平均绝对误差)、RMSE(均方根误差)和SSE(预测误差平方和)四种性能指标,最终得到结果为ARIMA-MSFD组合模型预测效果最优。

;In order to further improve the accuracy of power load prediction and make up for the limitations of single model,combined model has been widely used.Because there are many kinds of prediction models,it is particularly important to choose the right model for combination.Considering that the neural network may contain redundant independent variables and unreasonable network structure will seriously affect the learning accuracy and solving speed of the entire network,this paper uses ARIMA time series model and MSFD neural network model to combine and modify the predicted results.In order to verify the combination forecast model to get a higher forecasting accuracy,this paper used single ARIMA time series model,single MSFD neural network model and combined forecasting model of ARIMA-MSFD model for 2008 to 2018 in Gansu province on the hydraulic capacity for learning,and use this to forecast the 12-month output of electricity in 2019,by comparing three kinds of models to predict the results of MAPE(mean absolute percentage error),MAE(mean absolute error)and RMSE(root mean square error)and SSE(prediction error sum of squares)four performance indicators,The final result shows that the ARIMA-MSFD model has the best prediction effect.