基于改进BPNN算法的电力物资需求预测方法分析  

Analysis of Power Material Demand Prediction Method Based on Improved BPNN Algorithm

作　　者：张立波 刘俐君 ZHANG Libo;LIU Lijun(Taiyuan Power Supply Company of State Grid Shanxi Electric Power Company,Taiyuan,Shanxi Province,030001 China)

机构地区：[1]国网山西省电力公司太原供电公司,山西太原030001

出　　处：《科技资讯》2024年第15期44-46,共3页Science & Technology Information

摘　　要：电力物资需求预测是电力企业运营管理中一项重要技术,但是当前预测水平比较低,不仅平均绝对百分比误差(Mean Absolute Percentage Error,MAPE)比较大,而且预测结果置信度水平比较低,无法达到预期的预测效果,因此提出基于改进误差反向传播神经网络(Back Propagation Neural Network,BPNN)算法的电力物资需求预测方法。为保证电力物资需求预测的准确性,首先按照项目的性能将电力物资分为技改物资、维修物资、科技物资、基建物资以及信息物资,然后根据划分的物资类别从电力物资信息系统或平台上收集历史电力物资信息,并对数据泛化和归一化处理,利用改进BPNN算法对电力物资需求数据训练,提取电力物资需求特征,量化预测电力物资需求,以此实现基于改进BPNN算法的电力物资需求预测。经实验证明,设计方法MAPE不超过1%,预测结果置信度不低于95%,在电力物资需求预测方面具有良好的应用前景。Power Material Demand Prediction is an important technology in the application and management of power enterprises,but the current prediction level is relatively low.Not only MAPE is relatively large,but also the confidence level of the prediction results is relatively low,which cannot achieve the expected prediction effect.Therefore,a Power Material Demand Prediction method based on the improved BPNN algorithm is proposed.To ensure the accuracy of power material demand prediction,the power materials are firstly divided into technical materials,maintenance materials,scientific and technological materials,infrastructure materials and information materials according to the performance of the project.Then,according to the division of materials category,historical electricity material information is collected from the electricity material information system or platform,and the data is generalized and normalized.The improved BPNN algorithm is used to train data on power material demand,extract power material demand characteristics,and quantify the power material demand,so as to realize the power material demand prediction based on the improved BPNN algorithm.The experiment has proved that the design method MAPE is not more than 1%,and the confidence level of the prediction results is not less than 95%,which has a good application prospect in power materials demand prediction.

关 键 词：改进BPNN 算法 电力物资需求 电力物资信息系统 归一化 需求特征 量化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]