基于改进Elman神经网络的发电厂发电量预测

Power Plant Generation Prediction Based on Improved Elman Neural Network

为了减小计划发电量滞后性对发电厂生产活动的影响以及解决发电厂预测电量过程中缺乏科学指导的问题,提出一种结合数据重构和滚动预测的改进Elman神经网络的预测方法.采用变分模态分解(VMD)方法分解上海某电厂2016年-2022年的月度发电量数据,并建立基于天牛须搜索算法(BAS)改进Elman神经网络的发电量预测模型,将分解重构后的数据滚动输入预测模型,结果表明:VMD提高了数据的平稳性,经过分解、重构后的数据滚动输入模型,不仅扩大了原始数据量还避免了影响因素的分析,保留了原始数据特征,使模型的精度和预测速度有了显著提升,能够快速有效地指导发电厂制定燃煤采购及电力市场调度方案.

In order to reduce the impact of planned power generation lag on the production activities of power plants and solve the problem of lacking scientific guidance in the process of forecasting power generation,this paper proposes an improved Elman neural network prediction method combined with data reconstruction and rolling prediction,then uses the Variational Mode Decomposition(VMD)method to decompose the monthly power generation data of a power plant in Shanghai from 2016 to 2022,establishes a power generation prediction model based on the Elman neural network improved by the Beetle Antennae Search algorithm(BAS),and finally rolls the decomposed and reconstructed data into the prediction model.The results show that VMD improves the stationarity of the data,and that the model with data rolled in after decomposition and reconstruction not only expands the amount of original data but also avoids the analysis of influencing factors,with the original data characteristics retained,which significantly improves the accuracy and prediction speed of the model,and also helps to guide power plants to formulate coal purchase and power market scheduling plans quickly and effectively.