多应力下电能计量设备基本误差预估

Basic error estimate of electric energy metering equipment under multiple stresses

针对典型环境应力下电能计量设备基本误差受环境应力影响大,且多电应力间关系难以刻画的问题,提出一种改进粒子群长短时记忆网络(IPSO-LSTM)的电能计量设备基本误差预测方法。首先,对典型环境下多种应力数据进行归一化、数据集分配预处理;针对误差时序数据波动趋势,构建一种挤压LSTM网络结构以分析误差数据的变化趋势特征,以改善多应力数据下的模型非线性拟合能力;利用改进PSO算法对模型超参数进行寻优,减少模型超参数影响,增强模型预测效果。在实验部分,依据某公司的多个电能计量设备,结合新疆地区典型运行试验室测量的环境应力及其误差数据对所提出算法进行验证分析。结果表明,本文的样本预测精度指标RMSE分别达到1.08%和1.19%,MAE分别达到0.88%和0.96%。

The basic error of electric energy metering equipment is greatly affected by environmental stress. And the relationship between multiple electrical stresses hard to be described under typical environmental stress. To address these issues, an improved particle swarm with long short-term memory network(IPSO-LSTM) is proposed to predict the basic error of electric energy metering equipment. Firstly, various stresses data in typical environment are normalized and data set allocation are preprocessed. To solve fluctuation trend of the error time series data, an extruded LSTM network architecture is established to analyze the variation trend characteristics of the error data. In this way, the nonlinear fitting ability of the model under multiple stress data is enhanced. Then, the improved PSO algorithm is used to optimize the model′s hyperparameters to reduce the influence of hyperparameters and improve the prediction performance of the model. In the experimental part, the proposed algorithm is evaluated and analyzed according to several electric energy metering equipment of one company. The environmental stress and error data are both considered by typical operating laboratories in Xinjiang region. The results show that the sample prediction accuracy indexes RMSE values reach 1.08% and 1.19%, respectively. And MAE values reach 0.88% and 0.96%, respectively.