摘要
为有效保障智慧仓储系统的物资供应能力,对各类电力物资进行准确的需求预测是保证物资采购量和稳定高效供应的基础。针对现有智慧仓储系统接入数据少和难以支撑模型训练的问题,提出一种结合蒙特卡洛模拟和改进长短期记忆网络(long short-term memory,LSTM)的电力物资需求预测方法。首先根据初始数据集的分布和特征,采用蒙特卡洛方法模拟扩充数据集,同时利用KL(kullback-leibler)散度验证生成数据集的一致性,最后建立基于引导聚集算法的改进LSTM电力物资需求预测模型,提高模型的泛化能力和稳定性。通过仿真试验,所提模型有效提高了训练集可用数据过少前提下的电力物资预测精度。
In order to effectively guarantee the material supply capacity of the smart storage system,accurate demand forecasting of various power materials is the basis for ensuring material procurement and stable and efficient supply.Aiming at the problem of insufficient access data for existing smart storage systems and difficulty in supporting model training.It was proposed a method for power material demand forecasting that combines Monte Carlo simulation and improved long short-term memory(LSTM).Firstly,according to the distribution and characteristics of the initial data set,the Monte Carlo method was used to simulate the extended data set,and the Kullback-Leibler divergence was applied to verify the consistency of the generated data set.Finally,an improved LSTM power material demand forecasting model based on bootstrap aggregation(bagging)was established,so as to improve the generalization ability and stability of the model.Through simulation experiments,the model effectively improves the power material prediction accuracy under the premise of too little available data in the training set.
作者
陶加贵
孙毅
赵恒
管士宁
Tao Jiagui;Sun Yi;Zhao Heng;Guan Shining(Electric Power Research Institute,State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing Jiangsu 211103,China;School of Electrical and Electronic Engineering,North China Electric Power University,Beijing 102206,China)
出处
《电气自动化》
2023年第1期50-53,共4页
Electrical Automation
基金
国家电网有限公司科技项目“现代智慧仓储系统及智能物流装备关键技术研究”资助(1400-202118268A-0-0-00)。
