基于深度学习的SF_(6)气体压力时序数据预测

Time series data prediction of SF_(6) gas pressurebased on deep learning

为提高复杂电网环境中六氟化硫(SF_(6))气体压力预测的准确度,提出一种基于长短期记忆(long-short-term memory,LSTM)网络的SF_(6)气体压力预测方法。该算法根据SF_(6)气体压力实测数据对深度神经网络进行训练,并预测SF_(6)气体压力值,再将预测值与真实值的误差反馈给网络,以修正网络参数并提高预测准确度。仿真试验使用3个月的SF_(6)气体压力实测数据对LSTM神经网络进行训练和测试,结果表明,LSTM网络模型有效降低了SF_(6)气体压力预测误差。

To improve the accuracy of sulfur hexafluoride(SF_(6))gas pressure prediction in the complex grid environment,a long short-term memory(LSTM)network based SF_(6)gas pressure prediction method was proposed.According to the gas pressure data,the algorithm trained the deep neural network and predicted the SF_(6)gas pressure value,and then fed back the error between the predicted value and the true value to the network to correct the network parameters and improve the accuracy of the prediction.The simulation experiment used three months of SF_(6)gas pressure data to train and test the deep neural network.The results show that the LSTM network model can effectively reduce the prediction error.