Evaluation Method for Inaccurate Energy Measurement Based on Multi Source Measurement Data

A method for evaluating energy metering inaccuracies based on multi-source measurement data is proposed to address the issues of multiple types of non-equilibrium, measurement anomalies, and high cost of labeled samples in multi-source measurement data. Firstly, the feature data of the samples is extracted based on the encoder combined with attention mechanism, and the spatial distribution of the features of the electricity metering samples is obtained through contrastive learning method. Then set a group similarity threshold and calculate the average sample feature data of the group category. Subsequently, a complementary distance metric function constructed using cosine distance and Euclidean distance was used to evaluate the similarity of small sample learning classification. Finally, the complementary distance metric function is used to calculate the distance between the features of the mean grouped category samples, and feedback is provided to fine tune the encoder, contrastive learning, and softmax function parameters to achieve region oriented extraction of mean grouped category features. The experiment shows that in the process of inaccurate evaluation of electric energy measurement, the accuracy of the method proposed in this paper is not less than 98%, achieving timely detection and resolution of abnormal measurement data, improving the quality and availability of multi-source measurement data, and providing high-quality data support for the fusion application of measurement data at the provincial level.