一种基于信息熵的LSTM时间序列数据预测模型

时间序列预测可提升智能电网决策能耗评估有效性和电力传感网络的故障检测效率。基于香农信息熵和长短时记忆网络,构建一种基于时间序列数据的趋势预测模型,模型算法首先对时间序列数据以熵值法处理后进行特征归并,建立特征区间和熵值模型;其次在特征区间建立的基础上,将分类过后的数据在长短时记忆网络中进行训练得到预测结果。最后实验结果表明,与传统LSTM和GRU模型相比,高熵模型的均值平方差函数迭代结果误差降低85.9%和85.29%,显著改善模型预测结果的可靠性和准确性。

Time series prediction can improve the effectiveness of smart grid decision-making energy consumption evaluation and the fault detection efficiency of power sensor networks.Based on Shannon information entropy and long short-term memory network,a trend prediction model based on time series data is constructed.Firstly,the model algorithm merges the features of time series data with entropy method,and establishes the feature interval and entropy model.Secondly,on the basis of the establishment of the feature interval,the classified data are trained in the long-term memory network to get the prediction results.Finally,the experimental results show that,compared with the traditional LSTM and GRU models,the iterative error of the mean square variance function of the high entropy model is reduced by 85.9% and 85.29%,which significantly improves the reliability and accuracy of the model prediction results.