为提升电网公司计量器具管理水平，制定科学的物资采购计划，对计量器具需求预测的准确性及可靠性提出了更高的要求。文章以电能表为计量器具主要研究对象，依据简单循环单元（Simple Recurrent Unit, SRU）算法处理时序数据及并行计算的优势，以SRU算法为基础，结合径向基（Radial Basis Function, RBF）神经网络、贝叶斯网络对需求影响因素的数据处理，针对“业扩新装”、“轮换改造”及“故障抢修”需求类的计量器具分别构建需求预测组合模型，进行计量器具需求量的预测。实验数据表明，文章所提基于SRU组合模型的需求预测方法预测精度高且具有较高的运算效率，运用到电网公司业务中，经过验证该方法可为计量器具补货策略的制定提供有效的数据基础，进一步提高电网公司对计量器具的采购管理水平。
英文摘要:
In order to improve the management level of measuring instruments in power grid companies, and formulate a scientific material procurement plan, higher requirements have been put forward for the accuracy and reliability of demand forecasting of measuring instruments. The paper takes the electricity meter as the main research object of the measuring instrument. It is based on the advantages of the simple recurrent unit (SRU) algorithm to process time series data and parallel computing. Based on the SRU algorithm, it combines the radial basis function (RBF) neural network and Bayesian network to process the data of the demand influencing factors. And a demand forecasting combination model for “business expanding and new installation”, “rotational transformation” and “failure emergency repair” demand types of measuring instruments is established respectively, to forecast the demand for measuring instruments. The experimental data show that the combined forecasting model based on SRU proposed in this paper has high forecasting accuracy and high computational efficiency. It is applied to the business of power grid companies. The results show that the method can provide effective data basis for the formulation of replenishment strategy of measuring instruments and improve the purchasing management level of power grid companies.
